# RankVyze: expanded public reference

> RankVyze is a product launch directory and AI SEO platform that helps makers publish products and businesses improve their visibility in Google and AI search.

Canonical site: https://rankvyze.com
Contact: hello@rankvyze.com
Founder: [@iamumioz](https://x.com/iamumioz)
About: https://rankvyze.com/about

## Product launch directory

The RankVyze homepage at https://rankvyze.com/ is the primary product launch directory. The same published catalog can be read as JSON at https://rankvyze.com/catalog.json. Product details and claims may be supplied by makers. A listing is not an endorsement; editorial picks are labelled.

### Categories
- [AI products and tools](https://rankvyze.com/launches/categories/ai): Discover AI products for writing, research, automation, support and practical business workflows.
- [Productivity products and tools](https://rankvyze.com/launches/categories/productivity): Find productivity tools for planning, focus, notes, collaboration and everyday work.
- [Marketing products and tools](https://rankvyze.com/launches/categories/marketing): Explore marketing products for content, campaigns, analytics, research and customer growth.
- [Design products and tools](https://rankvyze.com/launches/categories/design): Discover design tools for visual work, prototyping, assets, interfaces and creative production.
- [Developer tools and products](https://rankvyze.com/launches/categories/developer-tools): Find developer tools for building, testing, deploying, monitoring and maintaining software.
- [SaaS products](https://rankvyze.com/launches/categories/saas): Discover software-as-a-service products for teams, creators and growing businesses.
- [SEO tools and products](https://rankvyze.com/launches/categories/seo-tools): Explore SEO tools for research, optimization, technical audits, reporting and AI search visibility.

### Published products
#### Notion
Listing: https://rankvyze.com/launches/notion
Category: SaaS
Pricing: See website
Maker or team: RankVyze editorial discovery
Organize knowledge, projects and work in one place.

Notion is a connected workspace for documents, knowledge, projects and collaboration, with AI features integrated into the product. Individuals and teams can shape pages and databases around their workflows. Visit the official website for current features and pricing.

#### Linear
Listing: https://rankvyze.com/launches/linear
Category: SaaS
Pricing: See website
Maker or team: RankVyze editorial discovery
Plan and build products with teams and agents.

Linear is a product development system for planning, tracking and building products. It brings initiatives, projects, issues and product workflows into a focused workspace designed for teams and AI-assisted work. Visit the official website for current features and pricing.

#### Railway
Listing: https://rankvyze.com/launches/railway
Category: Developer Tools
Pricing: See website
Maker or team: RankVyze editorial discovery
Deploy and operate software without cloud complexity.

Railway is a cloud platform for deploying applications and services from a repository. It provides automatic configuration, preview environments, networking, scaling, logs, metrics and rollbacks in one workspace. Visit the official website for current capabilities and pricing.

#### Supabase
Listing: https://rankvyze.com/launches/supabase
Category: Developer Tools
Pricing: See website
Maker or team: RankVyze editorial discovery
Build applications on a complete Postgres platform.

Supabase is a Postgres development platform that combines a database with authentication, data APIs, Edge Functions, realtime data, storage and vector support. Teams can use individual products or the integrated platform. Visit the official website for current features and pricing.

#### Claude
Listing: https://rankvyze.com/launches/claude
Category: AI
Pricing: See website
Maker or team: RankVyze editorial discovery
An AI assistant for thinking, creating and coding.

Claude is an AI assistant from Anthropic for writing, analysis, coding and working with documents and images. It is available on the web and mobile, with free and paid plans. Visit the official website to verify current capabilities, availability and usage limits.

#### Perplexity
Listing: https://rankvyze.com/launches/perplexity
Category: AI
Pricing: See website
Maker or team: RankVyze editorial discovery
Research the web with cited AI answers.

Perplexity is an AI answer engine that researches the open web and returns concise responses with citations. People can use it for questions, deeper research and analysis, while developers can build with its API. Visit the official website for current features and plan limits.

#### Cavyro
Listing: https://rankvyze.com/launches/cavyro
Category: Productivity
Pricing: See website
Maker or team: RankVyze editorial discovery
Keep customer relationships inside Telegram.

Keep customer relationships inside Telegram. Explore the product website for current features, availability and pricing.

#### CodeHype
Listing: https://rankvyze.com/launches/codehype
Category: SEO Tools
Pricing: See website
Maker or team: RankVyze editorial discovery
Help your SaaS get discovered across search and AI.

Help your SaaS get discovered across search and AI. Explore the product website for current features, availability and pricing.

#### DeskFerry
Listing: https://rankvyze.com/launches/deskferry
Category: Productivity
Pricing: See website
Maker or team: RankVyze editorial discovery
A simpler way to automate everyday business work.

A simpler way to automate everyday business work. Explore the product website for current features, availability and pricing.

#### Ditther
Listing: https://rankvyze.com/launches/ditther
Category: Design
Pricing: See website
Maker or team: RankVyze editorial discovery
An experimental playground for images and video.

An experimental playground for images and video. Explore the product website for current features, availability and pricing.

#### E-Warmup
Listing: https://rankvyze.com/launches/e-warmup
Category: Marketing
Pricing: See website
Maker or team: RankVyze editorial discovery
Build a healthier email sending reputation.

Build a healthier email sending reputation. Explore the product website for current features, availability and pricing.

#### NagMeLater
Listing: https://rankvyze.com/launches/nagmelater
Category: Productivity
Pricing: See website
Maker or team: RankVyze editorial discovery
Your reminders and daily tasks, right in WhatsApp.

Your reminders and daily tasks, right in WhatsApp. Explore the product website for current features, availability and pricing.

#### Poststack
Listing: https://rankvyze.com/launches/poststack
Category: Marketing
Pricing: See website
Maker or team: RankVyze editorial discovery
Plan and publish social content for your clients.

Plan and publish social content for your clients. Explore the product website for current features, availability and pricing.

## Published guides and articles

### Product launch alternatives

Written by RankVyze. Public platform references checked September 27, 2026; recommendations are editorial judgments, not controlled performance comparisons. Hub: https://rankvyze.com/alternatives

- [Product Hunt alternative](https://rankvyze.com/alternatives/product-hunt): A free launch listing, with a next step for AI visibility.
- [TinyLaunch alternative](https://rankvyze.com/alternatives/tinylaunch): Plan the work after the launch.
- [MicroLaunch alternative](https://rankvyze.com/alternatives/microlaunch): Give feedback a clear next step.
- [ScrollLaunch alternative](https://rankvyze.com/alternatives/scrolllaunch): Connect the listing to work on your own website.
- [StartupBase alternative](https://rankvyze.com/alternatives/startupbase): Choose a listing workflow that fits your release.

# Is AI search sending you customers? How to check in GA4

Canonical: https://rankvyze.com/blog/track-ai-referral-traffic-ga4
Published: 2026-09-28
Category: Guides

Find identifiable AI referrals in GA4, separate visits from brand mentions, and measure activation without treating all direct traffic as ChatGPT traffic.

Someone found your site through an AI answer. Great. Did they visit the pricing page, create an account, or disappear after three seconds? A citation screenshot cannot answer that. **Use GA4 to measure identifiable referral visits and the actions that follow. Use a separate research log to measure mentions and citations.** These are different datasets, and joining them by guesswork makes both less useful.

Start with the question you want to answer: which sources brought sessions that reached a meaningful outcome? You do not need a custom dashboard on day one. You need a source report, a landing page, and one event that means somebody got value. This guide covers the referral side; our [AI mentions and citations guide](/blog/measure-ai-mentions-and-citations) covers answer sampling.

## Start with Session source / medium

Open **Traffic acquisition** in GA4, set a date range, and change the table's primary dimension to **Session source / medium**. Search the table for the source you want to investigate. If the report is missing from your sidebar, an editor can restore it through the report library. Google's [Traffic acquisition documentation](https://support.google.com/analytics/answer/12923437?hl=en) explains the report and its session dimensions.

Look for observed source values such as `chatgpt.com` or `perplexity.ai`, rather than assuming every assistant sends an identical referrer. Those are examples to investigate, not a complete or permanent allowlist. Keep the medium beside the source. A manually tagged campaign and an ordinary referral from the same domain can otherwise look identical in your spreadsheet.

| Question | Useful dimension or metric | Avoid this shortcut |
| --- | --- | --- |
| Where did this visit come from? | Session source / medium with Sessions | Using the user's first acquisition source for every later visit |
| Which page started the visit? | Landing page, filtered to the same source cohort | Treating any page viewed later as the entry page |
| Did visitors do something valuable? | Your chosen key event and session key event rate | Calling every button click a customer |
| Was the brand recommended? | Saved AI answer and its wording | Inferring a recommendation from an analytics referrer |

## Build a small, auditable source list

Export the source rows you actually see. For each candidate, record the exact source value, medium, first observed date, and why you classify it as an AI referral. Keep a separate unclassified bucket. This gives you a report that can be explained to somebody else, instead of a regular expression that silently collects unrelated domains.

If you create a report filter or Exploration, begin with exact matches for confirmed values. Review new rows periodically. Do not include every domain containing the letters 'ai'. A broad filter can count your own newsletter, a directory, or an unrelated campaign. Likewise, a source value is attribution evidence, not proof of an individual person's identity or the exact prompt they used.

## Direct traffic is an unknown, not an AI category

Google explains that [`(direct) / (none)`](https://support.google.com/analytics/answer/15258820?hl=en) means Analytics lacks a clear referral source. Missing campaign parameters and lost referral information are among the possible causes. It does not tell you whether a visitor typed your URL, copied a link from an assistant, or came through another untracked path.

Keep direct traffic separate. If an identifiable AI referral count is small, report that count honestly rather than assigning a percentage of direct traffic to AI. You can add an optional 'How did you hear about us?' question after signup, but keep those answers as self-reported evidence. A customer may have seen you in three places before they registered.

> **A useful label:** Call the report 'Identifiable AI referral sessions'. That describes what the data can support. It is not a count of every AI mention, every AI-assisted visit, or all revenue influenced by AI.

## Measure activation before celebrating traffic

Choose one outcome that fits the product. For a project management tool, it might be creating a project and inviting a teammate. For a directory, it might be completing a valid product submission. A page view or scroll is easier to collect, but often says less about whether the visitor found what they needed.

The following numbers are a **hypothetical worked example**, not RankVyze performance data. Suppose identifiable AI sources bring 40 sessions and three sessions contain your activation event. The observed session activation rate is 3 divided by 40, or 7.5%. If one person fires the event five times in one session, that is still one activated session for this calculation. Event counts and session counts are not interchangeable.

Write the denominator beside the percentage. With only three successful sessions, the rate is fragile: a few additional visitors could change the picture substantially. Compare several consistent reporting periods before changing your budget. Also check whether internal tests, your own browsing, or repeated development events have entered the report.

## Use landing pages to choose the next improvement

1. **Find the entry page.** For the same AI source cohort, inspect the landing page. A tutorial and a pricing page serve different intentions; do not average away that distinction.
2. **Check the promise.** Read the page from a new visitor's perspective. Does it answer the likely question, and is the next action understandable without reading your whole site?
3. **Make one useful change.** Add a relevant example, clarify a limitation, or link to the next decision page. Record the change date and avoid redesigning five pages at once.
4. **Compare with context.** Review the next reporting period alongside traffic volume and campaigns. An association after an edit is not proof the edit caused the change.

For links you control, use a consistent campaign naming convention. Do not tag a directory link as 'ChatGPT' just because you hope an assistant will find it. Our [UTM guide](/blog/track-product-launch-traffic-utm) and [UTM builder](/tools/utm-builder) cover that separate job. RankVyze's [AI visibility workflow](/ai-search-visibility) can complement the report with recorded answer observations; it cannot recover referral data your analytics never collected.

### Can GA4 show the exact prompt someone typed into ChatGPT?
A normal source or referrer report does not provide the visitor's conversation. Do not infer an exact prompt from a landing page or source domain.

### Does zero AI referral traffic mean my brand never appears?
No. An answer may mention or cite you without generating a click. A click may also arrive without usable attribution, or without being measured by your analytics.

### Should I count all direct visits as AI traffic?
No. Direct is a lack of clear attribution. Keep it separate and report identifiable AI referrals as a limited, observable dataset.

### Sources and next steps
- [Google: Traffic acquisition report](https://support.google.com/analytics/answer/12923437?hl=en): Session dimensions and report navigation.
- [Google: traffic-source scopes](https://support.google.com/analytics/answer/11080067?hl=en): Why first-user and session acquisition answer different questions.
- [Google: direct traffic](https://support.google.com/analytics/answer/15258820?hl=en): What missing attribution does and does not mean.
- [Measure mentions separately](/blog/measure-ai-mentions-and-citations): Keep answer evidence separate from visit analytics.

---

# Google shows another company when I search my brand name. What now?

Canonical: https://rankvyze.com/blog/google-shows-another-company-for-my-brand-name
Published: 2026-09-28
Category: Technical

Separate an indexing problem from brand-name ambiguity. Audit your homepage, site name, canonical URLs, and public profiles with a practical evidence checklist.

You search your product name and find somebody else's business. The first instinct is to publish more pages. Before you do that, work out which problem you have. **A homepage that is not indexed, an indexed page that ranks poorly, and two businesses with similar names need different fixes.** Start with evidence about your own URL, then make your brand easier to distinguish.

This guide is about legitimate ambiguity, not reporting an unrelated company merely because it ranks above you. Search results can contain multiple businesses with similar names. The practical goal is to help people and search systems identify which website belongs to your product.

## Separate the three problems

| What you observe | What to inspect | First useful action |
| --- | --- | --- |
| Your homepage is absent even when you use the full URL | URL Inspection in Search Console | Check indexing status, access, and selected canonical |
| Your page appears for the domain but not the brand alone | Branded queries, page copy, and competing meanings | Improve identity consistency and relevant discovery links |
| Your result has an unexpected name above its title | WebSite markup and homepage naming | Check site-name signals separately from the title tag |
| An AI answer mixes facts from two companies | The cited sources and identifying details | Correct ambiguous source pages and record the wrong facts |

A `site:` query is a quick clue, not a complete index report. Inspect your canonical homepage URL inside your verified Search Console property. Record the indexed status, Google-selected canonical, last crawl information, and any access problem. A live test checks the current page; it does not itself mean the page is already indexed.

## Make one sentence do the identifying work

Imagine a fictional product called Harbor, a scheduling app for mobile repair teams. Its homepage says 'Work without limits'. That sentence could describe software, a coworking space, or an investment firm. A clearer opening is: 'Harbor is scheduling software for mobile repair teams, with technician calendars, job assignments, and appointment reminders.' It gives the name a category, audience, and concrete capabilities.

This is not a keyword-density exercise. Keep the brand spelling stable and describe what the product actually does. Use that same factual definition as the basis for the homepage, About page, and legitimate product profiles. Adapt the wording to each context; do not submit hundreds of identical paragraphs around the web.

Check your title tag too. A useful starting point for the fictional product is `Harbor | Scheduling software for mobile repair teams`. The visible heading and body should support that description. Google [generates title links from several sources](https://developers.google.com/search/docs/appearance/title-link), so the title you supply is an input rather than a guaranteed display string.

## Treat the site name and the page title as separate fields

The site name is the label that identifies the website in a search result. The title link describes the result page. Google's [site-name guidance](https://developers.google.com/search/docs/appearance/site-names) recommends `WebSite` structured data on the homepage, with a consistent preferred name and URL. An `alternateName` should be a genuine alternative used for the site, not a list of search keywords.

For Harbor, the site name could be Harbor while the homepage title adds the scheduling category. If Harbor Scheduling is a real public brand variant, it may belong in the naming signals. Do not invent a variant solely to impersonate an established company, and do not assume markup will make Google choose your preferred label immediately.

## Build an identity checklist before chasing backlinks

| Surface | Check | Example of avoidable ambiguity |
| --- | --- | --- |
| Homepage and About page | Same product name, category, and canonical domain | Homepage describes an app; About still describes a retired agency |
| Logo and visible wordmark | Consistent product spelling | Three spellings used across the header, footer, and images |
| Organization markup | Accurate name, URL, and genuine identifying profiles | sameAs points to an unrelated founder or uncreated account |
| Directory profiles | Correct official domain and product category | A typo sends visitors to a similarly named company |
| Old domains or URLs | Intentional redirects and consistent canonical signals | Two homepages claim to be the primary site |

Google says [Organization structured data](https://developers.google.com/search/docs/appearance/structured-data/organization) can help distinguish organizations. Use details you can substantiate. For `sameAs`, include only real profiles or pages that identify the same organization. Do not insert a future social profile, a directory homepage, or a competitor's URL. Leaving an unknown property out is better than creating a false relationship.

There is no need to publish a founder's personal information merely to fill every schema property. Represent the business accurately using the public information you intend to maintain. Structured data should agree with the page people can read, and it cannot replace a clear explanation of the product.

## Fix conflicting URLs without hiding the site

If HTTP, HTTPS, www, and bare-domain versions all exist, check which is intended to be canonical. Redirect genuine duplicates consistently, use the chosen version in internal links, and include that version in the sitemap. Google's [canonicalization documentation](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls) describes these signals. A canonical is not a method for claiming another company's pages.

Do not solve confusion by blocking your own homepage in robots.txt or adding `noindex`. Those controls can make discovery harder. If the issue is an old title, stale description, or retired brand wording, update the actual source first. Then request a recrawl for the changed URL rather than submitting unchanged URLs every day.

## Keep a weekly evidence sheet

1. **Record a baseline.** Save the exact query, date, country context, result URL, displayed site name, and title. Check both the brand alone and brand plus product category.
2. **Change the inconsistent sources.** Correct the homepage definition, title, canonical signals, and profiles you control. Keep a list of URLs changed and avoid inventing new facts.
3. **Request a recrawl where appropriate.** Use Search Console for a few changed URLs and keep the sitemap current. Submission is a request, not a deadline or ranking promise.
4. **Review actual discovery.** Look at branded-query impressions and clicks alongside the search examples. Distinguish a naming correction from a ranking improvement.

For AI answers, save the response and the linked sources separately. If an assistant says Harbor operates coworking spaces, the task is to identify where that fact came from, not to repeat 'Harbor scheduling' twenty times on the homepage. Our [guide to correcting wrong AI information](/blog/chatgpt-has-wrong-information-about-my-business) covers that workflow. A clear [product listing](/launch-your-product) can provide another consistent public reference, but no single listing guarantees brand rankings.

### Will Organization schema make my brand rank first?
No. It can communicate accurate identity information, but it does not guarantee a ranking, chosen site name, or knowledge panel.

### Should I rename my business immediately?
Not based on one search. First determine whether the issue is indexing, inconsistent identity, or genuine name ambiguity. A commercial naming decision needs more context than a search screenshot.

### How long does a branding correction take to appear?
There is no reliable fixed deadline. Search systems need to recrawl and process the affected pages, and the final display remains their choice.

### Sources and next steps
- [Google: site names](https://developers.google.com/search/docs/appearance/site-names): Site-name signals and homepage WebSite markup.
- [Google: Organization markup](https://developers.google.com/search/docs/appearance/structured-data/organization): Accurate organizational identity and disambiguation.
- [Google: canonical URLs](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls): Handle duplicate URLs consistently.
- [Diagnose an unindexed website](/blog/why-your-new-site-isnt-indexed): Resolve index eligibility before treating every absence as a branding issue.

---

# Your SaaS pricing page should answer more than 'How much?'

Canonical: https://rankvyze.com/blog/saas-pricing-page-ai-search
Published: 2026-09-28
Category: Strategy

Make SaaS pricing easier for buyers and search systems to interpret with clear billing units, worked totals, plan limits, accessible text, and consistent facts.

A pricing page can display a large '$19' and still fail to explain what the product costs. Is that per user? Per month? Only when paid annually? Does the customer need five seats? **A useful SaaS pricing page makes the billing unit, billing frequency, minimum commitment, and important limits explicit.** That helps buyers compare options and reduces ambiguity in the source material search systems can retrieve.

This is an editorial and usability checklist, not a trick for getting an AI recommendation. Google's [AI search guidance](https://developers.google.com/search/docs/appearance/ai-features) says ordinary SEO fundamentals apply to its AI features, including accessible text and structured data that matches visible content. Clear pricing does not guarantee that an assistant will read it, cite it, or describe it correctly.

## Write down the actual purchase equation

Before editing the page, ask someone on your team to calculate what a new customer would pay. Give them a realistic seat count and expected usage. If they need three separate help articles and a checkout preview to reach an answer, the pricing page is leaving too much work to the buyer.

Illustrative wording. Replace every value and condition with your actual offer.

| Pricing detail | Question the page should answer | Better than a bare number |
| --- | --- | --- |
| Billing unit | What does one unit buy? | $19 per user per month |
| Billing frequency | When is money collected? | Billed annually at $228 per user |
| Minimum commitment | How many units must I buy? | Minimum three users |
| Usage allowance | What is included and what counts? | 500 completed exports per workspace each month |
| Overage behavior | What happens after the allowance? | Exports pause until renewal unless you upgrade |
| Commercial conditions | What else can affect the total? | State applicable taxes, setup fees, or add-ons accurately |

## Use one worked total to expose hidden assumptions

Consider a **fictional** plan costing $19 per user per month when billed annually, with a three-user minimum. A three-person team pays $684 at the start of the year before any applicable taxes: 19 × 3 × 12. The monthly equivalent is $57 for that team, but the checkout charge is not $57.

Place that distinction close to the headline price. A visitor should not need to infer it from a tiny annual-billing toggle. If the product also offers monthly billing at a different rate, display both terms clearly. If a currency selector changes the offer, label the currency and relevant conditions in each state.

Do not describe an annual commitment as 'cancel anytime' without explaining what cancellation does. It might stop renewal while leaving the prepaid term active. Similarly, a refund guarantee needs the conditions that determine eligibility. The goal is to let a reasonable reader reconstruct the offer, not to bury the inconvenient parts below a bright CTA.

## Answer the comparison questions buyers actually have

Price alone rarely selects a plan. A buyer may ask whether a tool supports their existing platform, whether the free tier allows commercial use, or whether a teammate needs a paid seat just to review work. Give those questions a short, direct answer beside the relevant plan, then link to detailed documentation where needed.

| Buyer question | Useful answer pattern |
| --- | --- |
| Is there a free plan or only a trial? | State which exists, its duration if relevant, and the main restrictions |
| Which plan supports our workflow? | Name the required feature and the lowest plan that includes it |
| Can I export my data if I leave? | State the supported export format and any access limitations |
| Does this include implementation? | Separate software access from setup or managed service work |
| What does enterprise pricing depend on? | Explain the real quote inputs rather than inventing a starting price |

Avoid answering every question with 'Contact sales'. Custom pricing can be legitimate while still explaining what changes the quote: number of users, implementation scope, usage volume, or support requirements. Publish only factors that genuinely affect your offer. You do not have to reveal negotiated contracts to make the buying process understandable.

## Make the default HTML useful before interaction

Load the page without signing in and read what arrives before you touch a toggle. Can you identify the product, the plans, and their billing terms? Keep the essential explanation as text. A screenshot of a pricing table is hard to search, copy, translate, or inspect with assistive tools.

For interactive calculators, keep a plain-language explanation of the calculation alongside the interface. A calculator that initially shows '$0' while waiting for JavaScript may communicate the wrong offer if the surrounding text says nothing. Provide sensible labeled defaults and distinguish an estimate from a binding quote.

Test desktop and mobile states, annual and monthly selections, and the checkout destination. You are checking consistency across the purchase path. These checks do not require removing interactive elements; they require making the basic offer understandable without discovering every interaction.

## Treat schema as a description, not a discount machine

If you add `SoftwareApplication` or offer markup, keep it aligned with the visible software offer. Google's [software-app documentation](https://developers.google.com/search/docs/appearance/structured-data/software-app) describes its search-feature requirements. Schema.org vocabulary and eligibility for a Google rich result are not the same thing. Never invent reviews, ratings, or a free offer to satisfy a validator.

Validate what your CMS actually outputs. A template can retain last year's price after a designer updates the cards. Check the page source, any JSON-LD offers, the visible cards, and checkout together. Marking an amount as zero is not a substitute for saying 'contact sales', and a monthly equivalent should not lose the annual commitment that makes it available.

## Keep a small pricing source of truth

1. **Assign an owner.** Give one person responsibility for the public offer and record where prices are configured: billing provider, app, website, docs, and profiles.
2. **Change related surfaces together.** When an offer changes, update visible copy, structured data, help pages, and checkout. Retire outdated promotions or explain who still qualifies.
3. **Record real revisions.** Keep a dated internal change log. If you show a public update date, change it when the content is meaningfully revised, not on every deployment.
4. **Sample the questions again.** Check a small consistent set of pricing questions in the engines you care about. Save the date, answer, cited URL, and mistaken detail rather than simply recording 'wrong'.

> **Copy template:** [Product] offers [plan] for [amount and currency] per [billing unit], billed [frequency]. The minimum is [commitment]. This includes [main allowance]. After the limit, [actual behavior]. [Add accurate cancellation, trial, tax, or setup conditions where relevant].

If an assistant keeps quoting a retired plan, inspect the sources it links to. The stale fact may live in your own documentation, an old launch profile, or a third-party article. Update what you control and request corrections elsewhere. Our [wrong-information guide](/blog/chatgpt-has-wrong-information-about-my-business) explains that process, while the [product-page guide](/blog/product-pages-for-search-and-ai) covers the wider buying journey. RankVyze's [AEO service](/answer-engine-optimization) starts from those kinds of content and technical gaps, not a promise that markup can control an answer.

### Do I need public prices to appear in AI search?
There is no universal requirement to publish a fixed price. If you use custom quotes, explain the factors and process truthfully. Do not invent an amount for search engines.

### Will pricing schema stop AI tools quoting old prices?
No. Consistent markup can describe your current offer, but an answer can use another source or older information. Inspect the cited source when diagnosing a mistake.

### Should I remove my annual billing toggle?
Not necessarily. Keep the billing terms and default offer clear in accessible text, and verify every selection agrees with the checkout.

### Sources and next steps
- [Google: AI features and websites](https://developers.google.com/search/docs/appearance/ai-features): No special AI markup requirement; ordinary search fundamentals still apply.
- [Google: software application structured data](https://developers.google.com/search/docs/appearance/structured-data/software-app): Read the current feature requirements before implementing markup.
- [Write clearer product pages](/blog/product-pages-for-search-and-ai): Connect price to the buyer's actual decision.
- [Measure identifiable AI referrals](/blog/track-ai-referral-traffic-ga4): Evaluate visits and activation separately from answer visibility.

---

# A product launch SEO checklist: before, during and after launch

Canonical: https://rankvyze.com/blog/product-launch-seo-checklist
Published: 2026-09-20
Category: Guides

Prepare a product launch that people can find and use: a practical sequence for page readiness, search discovery, distribution and follow-up measurement.

A product launch has two jobs: help the right people understand the product today, and leave behind a useful destination they can find later. A burst of visits does not automatically accomplish the second job. This checklist connects the product page, the launch listing and the measurement plan so they support one another.

Start with the [free product launch checklist](/tools/product-launch-checklist). It contains 12 tasks and exports your progress as Markdown. It is a worksheet, not an automated audit. The workflow below explains what evidence to collect before you check each box.

## Before launch: choose one page and one user task

Pick the page you want launch visitors to land on. For a new product this may be the homepage; for a new feature it may be a dedicated feature page. Avoid announcing a specific capability and then sending everyone to a broad page where they have to search for it. Write down the action that shows the visit was useful, such as importing a first file, creating a workspace or requesting a relevant demo.

For an illustrative CSV cleanup product, the first action could be uploading a sample file and downloading a cleaned result. The page should explain supported formats, the file-size limit, what happens to uploaded data and whether an account is required. Those details are more useful than a broad promise to transform productivity. Show an actual workflow with a readable screenshot, and keep the screenshot consistent with the current interface.

| Check | Evidence to keep | Fix before announcing if… |
| --- | --- | --- |
| Product promise | One sentence describing user, task and outcome | A reader cannot tell what the product does |
| First useful action | A completed journey from a fresh account | Signup or the first task fails |
| Pricing and limits | A visible explanation beside the call to action | The visitor discovers an important restriction only after signup |
| Mobile usability | A test at a narrow viewport | The main action is obscured or requires horizontal scrolling |

## Check the page that crawlers and visitors receive

Inspect the deployed destination rather than a development preview. Check its response status, canonical URL, robots directives, title and description. Confirm that production has not inherited a staging noindex directive or login requirement. Run the [Meta Tag Checker](/tools/meta-tag-checker), [AI Crawler Checker](/tools/ai-crawler-checker) and [Sitemap Checker](/tools/sitemap-checker) to collect evidence, then inspect unexpected results directly.

Read the visible page as well. Important explanations should appear in readable text, with descriptive headings and useful links to setup instructions, pricing and support. A visitor should not need to interpret a screenshot to understand a limitation. If structured data is present, keep its claims consistent with the product people can actually see.

Link the destination from a relevant page on your own site. Add its canonical URL to the sitemap if it is intended for indexing. Google describes sitemaps and URL Inspection as discovery and recrawl mechanisms; neither guarantees inclusion. Use [Google’s recrawl guidance](https://developers.google.com/search/docs/crawling-indexing/ask-google-to-recrawl) when deciding between an individual request and sitemap submission.

## Launch day: distribute an accurate listing

Prepare a small listing kit: product name, short description, destination, category, screenshots and pricing summary. Keep the factual core consistent across placements, while explaining the use case in language appropriate to each audience. A developer community may care about API limits; an operations audience may care about setup and reporting. Do not copy a promotional claim you cannot support just because a directory field invites it.

Use the [UTM builder](/tools/utm-builder) for external campaign links when the platform permits them. Give each source a stable name and use a shared campaign name for the same launch. Open every final link after publishing. Redirects, misspelled domains and outdated destinations can break an otherwise good listing. If a platform removes your tags, record that limitation rather than treating untagged visits as zero interest.

- Record when each placement went live and save its URL.
- Check the listing on mobile and verify the call to action.
- Assign an owner to answer product questions and report bugs.
- Keep promotional statements factual; do not promise guaranteed rankings or AI recommendations.

## After launch: separate traffic, activation and discovery

Review outcomes in layers. First, did the links work and did people arrive? Second, did they complete the useful action you chose? Third, did they encounter a recurring obstacle? Search discovery and AI visibility belong in separate observations, because they are not interchangeable with visits. A product can receive a mention without a click, and a directory visit without becoming an activated user.

| Observation | What to inspect next |
| --- | --- |
| Many visits, few activations | Landing-page promise, onboarding steps and technical errors |
| Few visits from one listing | Audience fit, placement visibility and whether tracking survived |
| Repeated pricing questions | The placement of plan limits and pricing explanations |
| A new page is not indexed | Search Console evidence and the page’s technical state |

Set a review date rather than refreshing reports all day. A first-week review can identify broken experiences and unanswered questions; a longer observation window is needed for slower discovery changes. Save what you changed and why. This gives the next launch a useful baseline instead of a collection of impressions about what worked.

> **A completed checklist is not a ranking score:** It documents preparation. Customer demand, search inclusion, AI citations and conversions remain outcomes to observe.

### Sources and useful next steps
- [Launch readiness worksheet](/tools/product-launch-checklist): Check tasks and export an owner-ready handoff.
- [Launch directory](/launches): Explore products and prepare your own submission.
- [Google: request recrawling](https://developers.google.com/search/docs/crawling-indexing/ask-google-to-recrawl): Official options and their limitations.

---

# How to write a product directory listing people can actually evaluate

Canonical: https://rankvyze.com/blog/write-a-product-directory-listing
Published: 2026-09-29
Category: Guides

A field-by-field template for writing a product directory listing with a clear promise, proof, pricing context, useful visuals, and an honest call to action.

A good directory listing answers five questions quickly: what is this, who is it for, what job does it do, what does it cost, and what happens when I click through? Visitors are comparing several options at once. They should not need to decode a slogan to understand yours.

## Start with a sentence that can stand alone

Write the product name as it appears on your website. Then write a one-line pitch in this shape: **[Product] helps [specific user] do [specific job] by [distinctive method].** For example: “A planning workspace for small agencies that turns client requests into a prioritized weekly queue.” That is an illustrative example, not a claim about a real product.

Avoid “the future of productivity” or “AI-powered growth platform” when the reader cannot tell which task the product solves. If your category is broad, narrow the user or workflow. “For designers reviewing client feedback” is more useful than “for teams.”

## Give the visitor enough evidence to decide

- Explain the main workflow in plain language: what the user supplies, what the product does, and what they receive.
- Name two or three capabilities that distinguish the product. Prefer concrete behavior over adjectives such as seamless or revolutionary.
- State any important constraints: required integration, platform, account, region, or data access.
- Use a current logo and a legible screenshot of the actual product. Do not use a decorative mockup as a substitute for the interface.
- Describe pricing truthfully. If the product is free to start but paid for useful features, say so on the website users reach.

A listing is a discovery page, not a replacement for the product website. Give enough detail to make a qualified visit likely, then point people to the official site for current plans and capabilities.

## A field-by-field draft you can reuse

| Field | Weak draft | Useful draft |
| --- | --- | --- |
| One-line pitch | The next generation of work | Collect client feedback and turn it into prioritized tasks for a small agency |
| Description | We empower teams with AI | Upload a client brief, invite reviewers, and group comments by deliverable |
| Screenshot | Generic gradient illustration | Readable capture of the review queue, with sensitive data removed |
| Pricing | Affordable | Free trial; paid plans listed on the official pricing page |
| Call to action | Learn more | Try the review workflow on the official site |

The right category matters because it sets the comparison group. Pick the category that matches the core job, even if the product also uses AI. Add secondary context in the description rather than choosing a popular but misleading category.

## Check the listing after publication

1. **Review the live page.** Confirm the name, link, logo, screenshot, category, and pricing statement appear as intended on desktop and mobile.
2. **Test the destination.** Open the website link in a fresh browser session. It should reach the relevant product page without an unexpected sign-in wall or broken redirect.
3. **Measure useful visits.** Use a tagged URL only if the directory allows it, then compare qualified sessions, signups, and feedback—not only clicks. See the [launch UTM guide](/blog/track-product-launch-traffic-utm).
4. **Maintain the facts.** Update the listing when pricing, positioning, screenshots, or the product URL changes. An outdated directory result can confuse visitors and answer engines alike.

> **A practical final test:** Show the listing to someone outside your team for ten seconds. Ask them who the product serves and what it does. If they cannot answer, rewrite the opening two lines.

Ready to publish? [Submit a product to RankVyze](/launches/new), then use the [product launch SEO checklist](/blog/product-launch-seo-checklist) to prepare the destination page.

### How long should a product directory description be?
Long enough to explain the user, workflow, differentiation, and important constraints. Put the direct answer first; use the space the directory provides without filling it with repeated keywords.

### Should I reuse the same text in every directory?
Keep facts consistent, but adapt the opening and details to each directory's audience and fields. Maintain a canonical description internally so edits do not create contradictory claims.

### Sources and next steps
- [Google: Creating helpful, reliable, people-first content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content): Why original, useful detail matters more than keyword repetition.
- [RankVyze: Explore product launches](/launches): See how live listings are presented to visitors.

---

# Robots.txt allows AI crawlers. Why is Cloudflare still blocking them?

Canonical: https://rankvyze.com/blog/ai-crawler-blocked-by-cloudflare-waf
Published: 2026-09-24
Category: Technical

A diagnostic workflow for sites whose robots.txt permits OAI-SearchBot or PerplexityBot but whose CDN, WAF or bot challenge blocks real requests.

**A permissive robots.txt is only the first gate.** A CDN or WAF can still challenge or reject a crawler before it receives your page. If OAI-SearchBot is allowed in robots.txt but access logs show 403s, challenges or no successful fetches, inspect your edge security events and bot settings. Change the smallest rule that blocks a verified crawler; do not turn off protection for every bot.

This is a different diagnosis from [choosing which AI crawlers to allow](/blog/ai-crawlers-robots-txt). That guide covers crawler policy. This one covers delivery. Cloudflare [documents the interaction between AI Crawl Control and WAF rules](https://developers.cloudflare.com/ai-crawl-control/configuration/ai-crawl-control-with-waf/): an allowed AI crawler can still be blocked by an earlier WAF rule.

## Two gates to check, in order

| Gate | What to inspect | What a failure means |
| --- | --- | --- |
| Crawler policy | Production /robots.txt for the exact host and path | The crawler has been told not to fetch that path |
| Actual request | CDN security events and origin access logs | The request was blocked, challenged, redirected or served the wrong page |
| Delivered document | Final HTTP status and response body | A 200 can still contain a login screen, a JS challenge or no useful content |

Open the **production** robots file, not a local build or the www variant if the canonical host is bare. Confirm that a specific `User-agent` group and its path rules permit the page you want available. A broad `Allow: /` does not override a more specific applicable disallow under every possible configuration; use a robots tester or inspect the matching group. RankVyze's [AI Crawler Checker](/tools/ai-crawler-checker) can flag obvious policy problems, but it cannot prove that a real platform crawler passed your WAF.

## Find the block in security events

1. **Pick a page and a time window.** Choose one canonical URL, note its path, and inspect the CDN's security events for recent crawler requests to that host. Search for the relevant user agent, bot classification, rule ID, and request action.
2. **Verify the caller.** Do not trust a user-agent string by itself; anybody can copy it. Prefer your CDN's verified-bot classification or the platform's published verification method before changing a firewall rule.
3. **Read the action.** A block, managed challenge, JavaScript challenge or rate limit can stop a non-browser crawler. Check whether the event was caused by a custom rule, bot setting, AI Crawl Control, an origin firewall or a host-level access policy.
4. **Narrow the exception.** If the request is verified and you intend to allow it, scope the exception to that bot/category and necessary public paths. Keep private account, checkout and admin routes protected.
5. **Retest with real evidence.** Look for a later verified request that receives the intended page with a successful status. Save the event ID, rule changed and a sample response. Then monitor rather than assuming immediate inclusion in an AI answer.

Cloudflare's [bot reference](https://developers.cloudflare.com/ai-crawl-control/reference/bots/) distinguishes OAI-SearchBot, GPTBot, ChatGPT-User, Claude-SearchBot and PerplexityBot. These names do not all serve the same purpose. Cloudflare also documents a [known-bot field](https://developers.cloudflare.com/waf/custom-rules/use-cases/allow-traffic-from-verified-bots/) for custom rules. The exact controls available depend on your plan and configuration, so start from the event that blocked the request instead of copying a broad rule from a blog post.

## Why a curl test is not enough

You can run `curl -I -A 'OAI-SearchBot' https://example.com/page` to check what your server does with that string. It is a useful smoke test for redirects and accidental blocks. It is **not proof** that OpenAI can crawl you: the request comes from your machine, with a spoofed user agent and a different IP, bot score and challenge context. The reverse is also true: a 403 to your spoofed request does not prove the real verified crawler is blocked. CDN events for verified requests are stronger evidence.

> **A successful fetch is eligibility, not placement:** OpenAI says public pages can appear in ChatGPT search and recommends allowing OAI-SearchBot for summaries and snippets. It does not promise that an allowed page will be selected, cited or recommended.

## If the event log shows no crawler requests

Do not create an allow rule to solve a request you have not observed. Check whether the page is linked from discoverable pages, is publicly available without a login, has a stable canonical URL and actually returns useful text. Compare the relevant search engine's indexing status where one exists. Then record the absence as an observation, not proof that the crawler has blacklisted the site. Review our [indexing guide](/blog/why-your-new-site-isnt-indexed) and [AI rank checker](/ai-rank-checker) for separate discovery and answer checks.

### Does Allow: / in robots.txt bypass Cloudflare WAF?
No. Robots.txt states a crawling preference; it does not disable CDN or origin security rules. Inspect the CDN's event for the actual request.

### Should I allow every request claiming to be OAI-SearchBot?
No. User-agent strings can be spoofed. Use a verified-bot classification or the platform's published verification method and scope any exception narrowly.

### Will an allowed crawler make my site appear in ChatGPT?
Access is a prerequisite for OpenAI to read the page directly, not a ranking or citation guarantee.

### Sources and next steps
- [Cloudflare: AI Crawl Control and WAF order](https://developers.cloudflare.com/ai-crawl-control/configuration/ai-crawl-control-with-waf/): Official explanation of why an allowed AI bot can still be blocked.
- [Cloudflare: AI crawler reference](https://developers.cloudflare.com/ai-crawl-control/reference/bots/): Current crawler names and categories.
- [OpenAI: publishers and developers FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq): Crawler access and ChatGPT search eligibility.
- [Check your crawler policy](/tools/ai-crawler-checker): Start with the public robots file, then inspect real edge events.

---

# A 20-minute Search Console routine for the first weeks after launch

Canonical: https://rankvyze.com/blog/search-console-weekly-routine-after-launch
Published: 2026-09-29
Category: Guides

A weekly Search Console workflow to catch indexing problems, separate impressions from clicks, and choose useful product-page improvements after launch.

The first weeks after launch are noisy. Search Console can show whether Google can discover a page and whether it has begun appearing for queries, but a blank chart on day one is not a diagnosis. Run the same short review each week and record what changed.

## Before the first review

- Verify the correct domain property in Search Console.
- Submit a sitemap containing public canonical URLs, such as /sitemap.xml.
- Confirm your homepage and key product page are linked through normal navigation.
- Record the launch date, important URLs, and the main change you shipped.

A sitemap helps Google discover URLs; submission does not guarantee crawling or indexing. Google's documentation also says discovery and processing can take time. For a new site, use the URL Inspection tool on a few important URLs rather than treating every zero in a report as a failure.

## The 20-minute weekly review

1. **Minutes 0–5: Page indexing.** Check the Page indexing report for an unexpected rise in excluded pages. Prioritize important canonical pages. A filter URL or a duplicate that is excluded may be completely fine.
2. **Minutes 5–9: Inspect one or two key URLs.** For the homepage and product page, compare the user-declared canonical with Google's selected canonical. Check crawl status and whether indexing is allowed. Request indexing only after fixing a real issue or publishing a meaningful update.
3. **Minutes 9–14: Search performance.** Compare the latest complete week with the previous one. Split by pages and queries. Impressions with few clicks can suggest a weak title or mismatch between the snippet and the question; very low impressions may simply reflect a new site or weak demand.
4. **Minutes 14–17: Search appearance and devices.** Look for unexpected mobile or rich-result errors where applicable. Do not spend the week adding schema types your page does not truthfully support.
5. **Minutes 17–20: Choose one action.** Write down a single change with a reason: fix a blocked page, clarify a title, strengthen a thin product description, or add a relevant internal link. Review its effect next week.

## How to interpret the common patterns

| Pattern | Check first | Reasonable next step |
| --- | --- | --- |
| URL not indexed | Inspection status, canonical, robots, HTTP response | Fix the cause; then request indexing if appropriate |
| Impressions rising, clicks flat | Queries and title/snippet shown in results | Make title and opening answer more specific to the query |
| Clicks rising, signups flat | Landing page and source-to-signup journey | Improve the page's promise or call to action |
| A query appears for the wrong page | Internal links and page overlap | Clarify page purposes and link to the best answer |
| One-day spike or dip | Date range and annotation log | Wait for a fuller comparison before rewriting content |

Keep the site and the report in sync. A performance report can lag, and Search Console data is not a complete analytics system. Pair it with your own signup or conversion data. If your site is missing altogether, follow the [new-site indexing diagnosis](/blog/why-your-new-site-isnt-indexed) before rewriting titles.

## A simple review log

Use a spreadsheet with date, URL, observation, evidence, action, and next review date. Example: “Sep 29 /pricing — impressions for pricing queries, few clicks — title promises a free plan we do not offer — revise title and opening — review next Tuesday.” The example is a template, not RankVyze performance data.

For AI-search visibility, do not infer citations from Search Console clicks alone. Read the [AI mentions and citations measurement guide](/blog/measure-ai-mentions-and-citations) and keep separate evidence for those channels.

### Should I request indexing for every page each week?
No. Fix real access or content issues first. Repeated requests are not a substitute for crawlable internal links and a valid sitemap.

### When should I change a page title?
When query and page data show that the current title is inaccurate, unclear, or poorly aligned with the page. Preserve the page's purpose and compare a meaningful period afterward.

### Sources and next steps
- [Google Search Console getting started](https://developers.google.com/search/docs/monitor-debug/search-console-start): Official overview of performance, indexing, and inspection.
- [Google: Build and submit a sitemap](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap): What sitemap submission does and does not do.
- [Google: Title links in search results](https://developers.google.com/search/docs/appearance/title-link): Guidance for concise, descriptive titles.

---

# AI cites my website but never mentions my brand. What should I fix?

Canonical: https://rankvyze.com/blog/ai-cites-my-website-but-not-my-brand
Published: 2026-09-24
Category: Strategy

A page citation and a brand mention are different outcomes. Learn how to diagnose source attribution, add useful original evidence, and measure whether the name starts appearing.

**A source link is not a named recommendation.** If an AI answer links to your article but describes the information without naming your company, your page was cited in that answer; your brand was not necessarily endorsed or even mentioned. Treat those as separate outcomes. You can improve the clarity of who produced an original claim, but you cannot force an assistant to name you.

A [recent public discussion from a publisher](https://www.reddit.com/r/SEO/comments/1wesnrr/chatgpt_has_cited_about_700_of_our_pages_and/) describes exactly this frustration: many cited pages and very few named mentions. That is an example of the question people are asking, not proof of a universal citation-to-mention rate. RankVyze's [measurement guide](/blog/measure-ai-mentions-and-citations) defines both labels for a repeatable sample.

## First decide which outcome you need

| Outcome | Example observation | What it tells you |
| --- | --- | --- |
| Site citation | Answer links to your URL | That page was displayed as a source for this answer |
| Brand mention | Answer says your company name | The answer included the entity, with or without a link |
| Recommendation | Answer suggests your product for a buyer's need | The surrounding wording is positive and relevant to that decision |
| Referral visit | A person clicks through to your site | The answer led to a measurable visit, when tracking is available |

If your goal is referral traffic, a clearly labeled link may already help. If your goal is the brand being remembered, you need to measure named mentions and recommendation context. A branded prompt such as ‘What does Acme do?’ should be kept separate from an unbranded buyer prompt such as ‘Which tools clean CSV files for small teams?’ The first makes naming Acme part of the question and cannot serve as evidence that buyers would discover it unaided.

## Audit the cited page, not just your homepage

Open the exact URL the answer cited. Can a reader identify the publisher from the title, visible page header and article body? Is the most distinctive fact presented as an original observation with a method, date and named source? If a paragraph says only ‘many teams struggle with this’ then the answer may have little reason to name its author. If it presents a reproducible calculation or a documented case that your organization produced, attribution is meaningful to the human reader as well.

This is an editorial hypothesis to test, **not** a disclosed ChatGPT ranking rule. Improve the page because it becomes more useful and verifiable. Google explicitly asks whether content provides original information or analysis in its [people-first content guidance](https://developers.google.com/search/docs/fundamentals/creating-helpful-content). Do not invent a survey, source or credential to manufacture an attribution opportunity.

## Make the entity unambiguous where the claim lives

1. **Use a clear publisher name.** Put the current company name in the visible site header or article publisher area, and make sure it matches your About and contact pages.
2. **Name original work once, naturally.** If your team did the research, say who did it, what was measured, when it was measured and where a reader can inspect the method. Do not prefix every heading with the brand.
3. **Connect the topical page to the business.** Link from the article to a relevant product or method page and back when it helps the reader. A page about a problem should explain why your product belongs in the discussion, not just append a sales pitch.
4. **Keep structured data honest.** Organization and Article markup should reflect the visible publisher and page content. Google's [Organization guidance](https://developers.google.com/search/docs/appearance/structured-data/organization) says the markup can help disambiguate an organization; it is not a command to AI engines to mention it.

## Run a before-and-after test

Record a fixed set of unbranded prompts where the cited page is relevant. For each answer save the engine, date, visible settings, full answer and exact cited URL. Label site citation, name mention and recommendation separately. Publish one meaningful page improvement, then rerun the same question set on a schedule. A change in either direction is an observation; model variability, retrieval changes and competitor updates mean it is not automatically caused by your edit.

Use the [free AI Visibility Calculator](/tools/ai-visibility-calculator) to keep the numerator and denominator visible. If you want a single live Claude snapshot, the [AI rank checker](/ai-rank-checker) shows the answer and sources when configured. It does not stand in for ChatGPT, Gemini or Perplexity.

> **Avoid self-attribution spam:** Repeating the brand name next to every definition makes a page worse. Attribute only work your organization actually produced and keep the explanation useful even if a reader never becomes a customer.

### Does a citation mean ChatGPT recommends my product?
No. A citation means a page was linked as a source in that answer. Read the answer to determine whether the product was named, described accurately or recommended.

### Can schema force an AI assistant to name my brand?
No. Accurate publisher and Organization markup can help identify an entity in some search contexts, but it cannot guarantee a named mention.

### Sources and next steps
- [Google: helpful, reliable content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content): Primary guidance on original information, method and clear authorship.
- [Google: Organization structured data](https://developers.google.com/search/docs/appearance/structured-data/organization): What organization markup can and cannot identify.
- [Measure mentions and citations](/blog/measure-ai-mentions-and-citations): A repeatable observation method with separate denominators.

---

# IndexNow for product launches: what it sends, what it does not, and when to use it

Canonical: https://rankvyze.com/blog/indexnow-for-product-launches
Published: 2026-09-29
Category: Technical

Use IndexNow to notify participating search engines about a new or updated product page, verify ownership, and avoid confusing submission with indexing.

IndexNow is a notification protocol. You send a changed public URL to participating search engines so they can consider crawling it sooner. A successful response means the notification was accepted; it does **not** mean the page was crawled, indexed, or ranked.

## What IndexNow is useful for

- A new public product page that is linked from your site and appears in your sitemap.
- A meaningful update to a listing: name, description, pricing, or destination URL.
- A removed page that now returns an appropriate status or redirects to its replacement.

It is not a way to submit private dashboards, filtered duplicates, or a hundred unchanged pages every morning. And IndexNow is not Google's URL submission mechanism. Keep a valid sitemap and use Search Console for Google-specific diagnostics.

## The setup in four steps

1. **Create an ownership key.** Generate a key following IndexNow's documented format and put the matching text file at the root of your host. The file proves you control the site.
2. **Publish the canonical URL.** Ensure the page returns 200, is public, allows indexing, uses the intended canonical URL, and is linked internally. A notification cannot repair a blocked or duplicate page.
3. **Notify the endpoint.** Send the URL or a small batch using the documented API format, including host, key, keyLocation, and urlList as needed. Send only URLs under the verified host.
4. **Verify independently.** Check the API response, then inspect the URL in Bing Webmaster Tools. Later, check whether it was actually crawled and indexed. These are separate events.

## An example request

Illustrative JSON body for a batch notification. Replace the host, key, and URL with your own verified values.

```json
{
  "host": "example.com",
  "key": "YOUR_OWNERSHIP_KEY",
  "keyLocation": "https://example.com/YOUR_OWNERSHIP_KEY.txt",
  "urlList": ["https://example.com/products/new-tool"]
}
```

For a product launch, send the canonical product URL after the page is live. If you also publish an announcement, that distinct article may be submitted too. A tracking URL with utm_source is usually not the canonical page to send; keep campaign measurement separate from indexable destinations.

## What to check when nothing appears

| Symptom | Likely check |
| --- | --- |
| API rejects the request | Key file accessibility, exact key match, URL host, request syntax |
| Accepted but not crawled | Internal links, sitemap, server availability, crawl controls |
| Crawled but not indexed | Canonical selection, duplicate content, page usefulness, noindex |
| Indexed but no impressions | Demand, title relevance, competition, and time |

RankVyze uses an IndexNow key file and a submission script that checks its live sitemap before sending changed URLs. That is a safety guard, not a promise that search engines will index every launch. For the full launch sequence, see the [product launch SEO checklist](/blog/product-launch-seo-checklist) and [canonical campaign URL guide](/blog/canonical-urls-for-launch-campaigns).

### Does IndexNow submit my page to Google?
No. IndexNow is used by participating search engines. For Google, use crawlable links, a sitemap, and Search Console to inspect important URLs.

### Should I send every URL every day?
No. Notify new, updated, or removed URLs. Repeatedly submitting unchanged URLs does not create useful content or guarantee indexing.

### Sources and next steps
- [Bing: IndexNow getting started](https://www.bing.com/indexnow/getstarted): Ownership key, endpoints, and request format.
- [Google: Build and submit a sitemap](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap): Google's discovery and sitemap guidance.
- [RankVyze: Why a new site may not be indexed](/blog/why-your-new-site-isnt-indexed): Diagnose crawl and indexing problems separately.

---

# ChatGPT has wrong information about my business. How do I correct it?

Canonical: https://rankvyze.com/blog/chatgpt-has-wrong-information-about-my-business
Published: 2026-09-24
Category: Guides

A source-first workflow for correcting outdated product descriptions, prices and brand confusion in AI answers without pretending a website edit controls a model.

**You cannot directly edit an ordinary ChatGPT search answer about your business.** You can correct the public information it might retrieve, identify outdated third-party pages, and check whether a new answer uses the corrected facts. Start by saving the exact wrong statement and its visible sources; changing a homepage without knowing where the bad fact came from is guesswork.

A [marketer's public question](https://www.reddit.com/r/SEO/comments/1vblt24/anyone_else_checking_how_chatgpt_describes_their/) describes the common problem: the same company is represented differently across AI tools. That is a useful prompt for investigation, not evidence that every answer uses the same sources. OpenAI itself notes that [search results and citations can be incomplete, outdated or incorrect](https://help.openai.com/en/articles/9237897-chatgpt-search).

## Classify the error before you try to fix it

| Error | Likely place to investigate | Useful next action |
| --- | --- | --- |
| Old price or feature | Your pricing/docs and cited directories | Update the canonical product page and request correction on old listings |
| Confused with a similarly named business | Search results, titles, About page, external profiles | Clarify the entity and domain wherever the wrong identity appears |
| Invented claim with no matching source | Answer text, cited URLs, search mode | Document it, retest in a fresh session and avoid repeating it as a fact |
| Correct fact but outdated answer | Page update date and cached or cited page | Verify the current page is crawlable, then allow for retrieval and recrawl delays |

## Trace the answer back to evidence

1. **Preserve one reproducible example.** Save the full prompt, date, answer, browsing/search state and source links. Note which words are wrong. Do not rely on a cropped screenshot that hides citations or context.
2. **Open every cited page.** Look for the wrong statement in the page body, snippet and page title. Record the exact URL. A cited page may support only part of the answer, so do not assume that every sentence came from it.
3. **Check your own canonical facts.** Compare the public homepage, product page, pricing, documentation, About page and metadata. If they disagree, correct the authoritative page first and update or redirect obsolete pages.
4. **Correct outside sources you control.** Update directory profiles, product listings and social profiles. For independent articles, ask the publisher for a factual correction with the current source URL. Do not buy or fabricate reviews to overwhelm the error.
5. **Retest carefully.** Ask the same question in a fresh conversation with search enabled where available. Record the new answer and sources. A corrected page does not guarantee a different answer or a specific update time.

## Write a page that answers the disputed question

If the recurring error is ‘Company X only works with large enterprises,’ put the real eligibility on a relevant public page in plain HTML: who the product is for, plan limits, supported integrations and a date if facts change. A vague hero slogan does not settle a specific question. Include concrete, independently checkable facts and link to the documentation that explains them.

Use one stable business name and domain in the visible site header, About page, contact information and relevant listings. If another organization uses a similar name, describe the category and product clearly. Google's [site-name guidance](https://developers.google.com/search/docs/appearance/site-names) and [Organization markup guidance](https://developers.google.com/search/docs/appearance/structured-data/organization) explain how consistent names and accurate `sameAs` profiles can help disambiguate a website in Google Search. They do **not** promise to overwrite an AI model's answer.

## What if the answer has no visible citation?

Do not claim to know its internal source. Search for distinctive phrases from the false statement, inspect common directories and old press pages, and check whether your own site still says something similar. Correct verified errors you find. If no source can be identified, keep the question in a monitoring set and label its source as unknown. OpenAI notes that models can produce inaccurate answers; a missing citation is not permission to invent a causal explanation.

> **Personal-data complaints are a different route:** If the incorrect answer concerns a person's private information rather than business positioning, use the relevant platform's privacy or correction process. Editing public marketing copy is not a substitute for that process.

### Can I force ChatGPT to update my company description?
No general website tag or request guarantees that outcome. Correct public sources, make the facts easy to retrieve and retest; answer generation remains outside your direct control.

### Should I add a sameAs link to every profile?
Only add genuine profiles that describe the same organization. Invented or unrelated profiles do not help users verify the entity.

### How fast do corrected facts appear?
There is no reliable universal timetable. Crawl, retrieval, model behavior and the question asked all affect the next answer.

### Sources and next steps
- [OpenAI: searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-chatgpt-search): Official explanation of sources and answer limitations.
- [Google: site names](https://developers.google.com/search/docs/appearance/site-names): Consistent naming and site-level identity signals.
- [Google: Organization markup](https://developers.google.com/search/docs/appearance/structured-data/organization): Accurate organization identity and sameAs guidance.
- [Check whether ChatGPT knows your business](/blog/how-to-check-if-chatgpt-knows-your-business): A separate observation workflow for current answers.

---

# Are product directory links useful for SEO? A practical evaluation guide

Canonical: https://rankvyze.com/blog/product-directory-links-seo
Published: 2026-09-20
Category: Strategy

Evaluate product launch directories by audience fit, listing quality, referral outcomes and transparent link policies—not promises of guaranteed rankings.

A product directory can help a buyer discover a tool, compare alternatives or understand a category. That is a concrete reason to publish a useful listing. The SEO value of an individual link is much harder to isolate, and a promise of guaranteed ranking improvement is a poor basis for choosing where to spend time or money.

Use this guide to evaluate a directory as a distribution channel. RankVyze operates a [product launch directory](/launches), so we have a direct interest in the category. Apply the same checks to our listings and to every other platform you consider. This is an evaluation method, not a ranking of competing directories.

## Start with the visitor, not the link metric

Describe the person you want to reach and the decision they are making. A database developer selecting a monitoring tool has different needs from a designer browsing inspiration. A large general directory may expose a product to many people but offer little context for a specialized purchase. A smaller relevant community may be easier to serve with a clear demo and a thoughtful answer to a question.

Before submitting, browse the actual category where your product would appear. Read several recent listings. Are descriptions specific enough to distinguish products? Do links lead to working destinations? Can a visitor understand pricing and limitations? Can the listing be updated after the product changes? These observations are more actionable than a third-party authority number alone.

| Criterion | Evidence to inspect | Why it matters |
| --- | --- | --- |
| Audience fit | Relevant categories and realistic buyer questions | Exposure only helps if the visitor has a reason to care |
| Editorial quality | Accurate listings, moderation and corrections | Useful context helps visitors make a choice |
| Destination quality | Working links and transparent redirects | The click must reach the promised product |
| Commercial terms | Clear fee, duration and paid-placement labels | A placement should have a defined deliverable |
| Measurement | Permitted campaign links or referral reporting | You need a way to learn from the placement |

## Understand what a link attribute does and does not say

Google’s [outbound-link guidance](https://developers.google.com/search/docs/crawling-indexing/qualify-outbound-links) describes sponsored for paid placements, ugc for user-generated links and nofollow for certain other relationships. Inspect the actual link when its treatment matters to your decision. A label in a sales page is not a substitute for what the published page serves.

A qualified link can still be useful to a person who clicks it. Conversely, an unqualified link is not proof that a placement will improve rankings. Treat the link relationship as one fact about the listing, alongside its audience, wording, placement and cost. Do not ask an editor to hide sponsorship or remove a required qualification to make a paid placement look editorial.

Google’s [spam policies](https://developers.google.com/search/docs/essentials/spam-policies#link-spam) identify low-quality directory links and link purchases intended to manipulate rankings among link-spam examples. This is a reason to avoid packages whose main deliverable is a large count of links. It does not make every directory or paid advertisement inherently useless; purpose, quality and qualification matter.

## Run a small, measurable distribution experiment

Choose a few relevant placements and write down a hypothesis for each. For example: ‘Operations teams browsing this category may try our CSV cleanup workflow.’ Decide what would count as a useful outcome before seeing the results. That might be qualified enquiries, completed imports or feedback from the intended audience. A raw signup can be a weak measure if most accounts never use the product.

Use consistent [campaign links](/tools/utm-builder) where allowed, record publication dates and keep the landing-page offer reasonably stable during the review window. If one placement runs a paid newsletter and another provides only a permanent profile, compare the deliverables honestly. They are different distribution formats, even if both include a link.

Invented numbers for a calculation example, not RankVyze or customer performance.

| Illustrative placement | Visits | Activated users | Cash cost | Observed cost per activation |
| --- | --- | --- | --- | --- |
| Listing A | 120 | 6 | $60 | $10 |
| Listing B | 30 | 3 | $0 | $0 cash; staff time excluded |

In this example A produces more activations, while B produces fewer at no cash charge. Neither conclusion tells you which users will retain or pay. Add the time needed to prepare and maintain the listing, and follow the activated cohort long enough to assess its quality. Do not label these observational results incremental lift without a suitable experiment.

## Maintain the listing after the announcement

A directory profile is a small product page on another site. When pricing, availability or positioning changes, update it. Remove broken screenshots and replace unsupported claims. A clear listing can answer basic questions before the visitor reaches your site; an outdated one can set the wrong expectation before you have a chance to explain the product.

Make a simple inventory with the listing URL, owner, last review date and destination. Review it alongside your product documentation. If a platform cannot correct a material error or insists on a misleading claim, reconsider whether the exposure is worth keeping. The aim is a reliable path to your product, not a badge collection.

### Should I submit to every directory I can find?
Prioritize relevant audiences and maintainable listings. A larger submission count alone does not demonstrate useful distribution or SEO value.

### Does a nofollow link have no value?
It may still send interested visitors and help them discover a product. Do not equate referral value with a guaranteed search-ranking contribution.

### Sources and useful next steps
- [Google: link qualifications](https://developers.google.com/search/docs/crawling-indexing/qualify-outbound-links): Definitions for sponsored, ugc and nofollow.
- [Google: spam policies](https://developers.google.com/search/docs/essentials/spam-policies#link-spam): Primary guidance on manipulative link practices.
- [Track a launch campaign](/blog/track-product-launch-traffic-utm): A reporting workflow for external placements.

---

# SoftwareApplication schema for SaaS pages: a truthful implementation checklist

Canonical: https://rankvyze.com/blog/softwareapplication-schema-for-saas
Published: 2026-09-29
Category: Technical

A practical guide to SoftwareApplication structured data for SaaS pages, including eligible fields, honest pricing, validation, and limits of rich results.

SoftwareApplication structured data can help describe a software product to search engines. It is not a switch that makes a SaaS appear in AI answers, and valid JSON-LD does not guarantee a rich result. Start with a useful, crawlable product page; mark up facts that people can see there.

## Decide which page owns the product facts

Use the official product page as the primary source for current features, platforms, and pricing. A directory listing can describe the product, but it may not know every plan or recent change. Avoid publishing conflicting prices across pages. If a directory cannot keep pricing current, linking to the official pricing page can be more honest than inventing a price.

## Fields to verify before implementation

| Field | What to check |
| --- | --- |
| name | Matches the visible product name |
| applicationCategory | Accurately describes the software; use Google's supported guidance |
| operatingSystem | Only platforms the product actually supports |
| offers | The visible price and currency match the offer; do not label a paid product free |
| aggregateRating | Only genuine ratings displayed on the page and permitted by the applicable rules |
| url and image | Resolve to the canonical product page and a real, accessible image |

Google's software-app rich-result documentation lists required and recommended properties. Requirements for a particular appearance can change, so check the current documentation and test the final page, not a copied snippet in isolation. An entity description and rich-result eligibility are different questions.

## A minimal example to adapt

Illustrative data only. Remove or change any property that does not match your visible page and current offer.

```json
{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "Example Task Planner",
  "applicationCategory": "BusinessApplication",
  "operatingSystem": "Web",
  "url": "https://example.com/product",
  "offers": {
    "@type": "Offer",
    "price": "12.00",
    "priceCurrency": "USD"
  }
}
```

If a product has multiple plans, a single number may mislead. Check the applicable Google guidance for supported offer structures and make the visible pricing explanation clear. Never fabricate reviews, availability, or a zero-dollar price merely to satisfy a validator.

## Validation workflow

1. **Compare against the page.** Read each marked-up statement beside the visible product copy. If a visitor could not verify it, remove or explain it.
2. **Check syntax and eligibility.** Run Google's Rich Results Test and inspect the extracted item, warnings, and errors. Syntax validity alone does not prove that a result will be shown.
3. **Publish and inspect.** Deploy the page, check its canonical and robots state, then use Search Console URL Inspection. Monitor enhancement reports if Google provides them for the markup.
4. **Maintain it.** Review markup whenever pricing, supported platforms, branding, or the primary URL changes.

> **AI-search reality:** Google says its AI features rely on ordinary Search fundamentals and do not require special AI-specific schema. Use structured data to state real facts clearly, while improving the actual page remains the main job.

For broader entity and markup choices, read the [schema and AI-search guide](/blog/schema-markup-for-ai-search). If you are preparing a directory listing, the [listing writing template](/blog/write-a-product-directory-listing) covers the visible copy that should come first.

### Will SoftwareApplication schema make my SaaS appear in ChatGPT?
No such outcome is guaranteed. Accurate, accessible product information is useful, but schema by itself does not cause an AI engine to mention or recommend a product.

### Can I add a five-star rating without reviews on the page?
No. Do not invent rating data or mark up reviews visitors cannot verify. Follow the current search engine requirements for review snippets.

### Sources and next steps
- [Google: Software app structured data](https://developers.google.com/search/docs/appearance/structured-data/software-app): Current properties, examples, and eligibility requirements.
- [Google: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features): Ordinary indexing and content requirements for AI search features.

---

# I rank on Google, so why don't AI answers cite my page?

Canonical: https://rankvyze.com/blog/rank-on-google-but-not-cited-by-ai
Published: 2026-09-24
Category: Strategy

Diagnose why a Google-ranking page is absent from ChatGPT or AI Overview sources: eligibility, question fit, answer structure, source selection and measurement.

**Ranking in organic search does not reserve a citation in an AI answer.** First identify the surface: Google AI Overviews require a page to be indexed and eligible for a snippet, while ChatGPT search has its own crawler and source-selection process. Then check whether the tested question actually calls for your page's information. An unrelated page-one ranking for a broad keyword is weak evidence that your page should answer a more specific buyer question.

Google's [AI-features documentation](https://developers.google.com/search/docs/appearance/ai-features) says there are no special technical requirements beyond Search eligibility for an AI Overview supporting link, and that inclusion is not guaranteed. OpenAI's [publisher FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq) recommends allowing OAI-SearchBot for content to be included in ChatGPT summaries and snippets. These are separate systems; do not apply a Google indexing rule to ChatGPT or vice versa.

## Use the right diagnostic branch

| Observation | First check | Do not conclude |
| --- | --- | --- |
| Google page one, no AI Overview appears | Whether an AI Overview triggers for the exact query, market and moment | That your site failed an AI-specific technical test |
| AI Overview appears, your URL absent | Google index and snippet eligibility; relevance of page to the full question | That schema markup alone will fix selection |
| ChatGPT names a competitor, your page absent | OAI-SearchBot access, question fit and the sources ChatGPT actually linked | That Google position controls ChatGPT's answer |
| AI cites your URL but does not name you | Brand attribution in the answer and the cited page | That citation equals recommendation |

## For Google AI Overviews: verify eligibility first

Inspect the exact canonical URL in Google Search Console. Confirm that it is indexed and eligible to appear with a snippet. Check for noindex, nosnippet, blocked crawling, redirects, a surprising Google-selected canonical and a page that delivers little text to Googlebot. [Google's URL Inspection help](https://support.google.com/webmasters/answer/9012289) explains what the indexed and live tests can show. Fix a real technical fault before writing another ‘AI optimized’ variant of the same page.

If the page is eligible, look at the full question. Google says AI features may use [multiple related searches](https://developers.google.com/search/docs/appearance/ai-features) to assemble an answer. A page that defines ‘CRM’ might rank for that term but have little to say about ‘Which CRM supports offline field sales for a five-person team?’ Do not add a generic FAQ just to match the phrase. Add a useful, supported answer if your product or evidence genuinely serves that use case.

## For ChatGPT: inspect access and answer relevance separately

Read the production robots policy for OAI-SearchBot and, if you can, inspect verified crawler requests at the CDN. A robots allow rule is not proof of delivery: [WAF rules can still block an allowed AI crawler](/blog/ai-crawler-blocked-by-cloudflare-waf). Then save the actual answer and its sources. A competing page may answer a different sub-question, contain more specific evidence or be better matched to the user's context. The source list is an observation about that answer, not a complete ranking report.

## Rewrite for a question, not a keyword clone

1. **Choose one real buyer question.** Use the precise decision a visitor is trying to make, including audience, constraints and comparison criteria. Write it down before editing.
2. **Identify the missing answer.** Read your current page and the cited sources. What fact or method does the answer need that your page lacks? If your page already answers it well, keep the content and test other questions rather than rewriting for a single result.
3. **Add verifiable substance.** Give a direct answer followed by limits, examples, comparisons, dates or a reproducible method. Distinguish facts about your product from broader claims requiring external evidence.
4. **Link it into the site.** Connect the improved page to relevant product and explanatory pages with descriptive internal links. Keep one canonical page per distinct intent instead of publishing many near-identical variants.
5. **Measure the next sample.** Retest a fixed question set and record mention, citation and recommendation separately. Note engine, date and exact URL so a different answer can be investigated.

For a concrete example, a scheduling tool may rank for ‘meeting scheduling software’ with a feature list. A buyer asks an AI assistant for a tool that handles round-robin routing across two time zones without requiring a paid seat for every viewer. A useful page would explain how that exact workflow works, its limits and current pricing. A paragraph repeating ‘best meeting scheduling software’ ten times does not answer the question.

> **No special AI markup:** Google says you do not need new machine-readable files or special schema to appear in its AI features. Keep structured data accurate and consistent with visible text; focus first on eligibility and a genuinely useful answer.

### Does a page-one Google result have to appear in AI Overviews?
No. An AI Overview may not appear for the query, and an indexed, snippet-eligible page is eligible rather than guaranteed to be selected as a supporting link.

### Will more FAQ schema force a citation?
No. Google says no special schema is needed for AI features. Use only markup that accurately describes visible content and meets applicable guidelines.

### Can a page rank on Google while ChatGPT cannot crawl it?
Yes. Google and OpenAI use different crawlers, and an edge rule may treat them differently. Check each access path separately.

### Sources and next steps
- [Google: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features): Indexing, snippet eligibility and no special AI requirements.
- [OpenAI: publishers and developers FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq): ChatGPT search crawling and inclusion guidance.
- [Google: URL Inspection](https://support.google.com/webmasters/answer/9012289): Check the indexed version and live page.
- [Build a product page that answers buyers](/blog/product-pages-for-search-and-ai): A deeper page-writing workflow.

---

# How to measure AI mentions and citations without inventing a visibility score

Canonical: https://rankvyze.com/blog/measure-ai-mentions-and-citations
Published: 2026-09-20
Category: Guides

Define a prompt sample, count brand mentions and site citations separately, and compare AI visibility observations with clear denominators and limitations.

An AI answer can name your product without linking to it. It can link to a page on your site without spelling out the brand. It can also mention the brand while explaining why it is unsuitable. Counting all three as a single success hides the difference between being present, being cited and being recommended.

This guide proposes a manual observation method you can reproduce with a spreadsheet and the [AI Visibility Calculator](/tools/ai-visibility-calculator). The formulas describe your sample only. They are not an industry-standard ranking score, an estimate of all AI users or evidence that a website edit caused a change.

## Define the question set before you collect answers

Start with the decisions your intended customer needs to make. For an illustrative meeting-notes product, useful questions might concern multilingual calls, exporting action items or comparing tools for a small team. Keep questions that include your brand in a separate group. Asking an engine directly about your product makes a mention much more likely than an unbranded selection question does.

Write down a fixed set of prompts and label their intent. Keep engines separate and record whether browsing was enabled. Where visible, preserve the model or product version, date, location, account context and whether the conversation was new. These details do not remove variability; they make it easier to understand what changed between runs.

| Field | What to record |
| --- | --- |
| Prompt ID and exact wording | A stable identifier plus the complete question |
| Engine and settings | Product, visible model, browsing state and relevant account context |
| Time and location | Date, time zone and market being tested |
| Answer evidence | Saved answer text and the displayed source URLs |
| Outcome labels | Mention, site citation, accuracy and recommendation context |

## Use definitions another reviewer can follow

For this method, a mention means the answer names the brand or an agreed, unambiguous variation. A citation means the answer includes a link to a URL on the site being measured. Count each answer at most once for each metric. Three links to your domain in one answer still produce one cited answer. A third-party review about your brand is a different category from a citation to your own site.

Decide how to handle ambiguous product names before scoring. If your brand shares its name with a common word, require enough context to identify it. Add an uncertainty column rather than forcing every ambiguous reference into yes or no. Have a second person review disputed cases if you can, and preserve the explanation for the final label.

Track unsuccessful requests separately. A network error is not an answer. A completed response that says it cannot help may reasonably remain in the sample under a predefined rule. The important point is to document exclusions and keep the same rule across periods; otherwise the denominator can quietly change to make a report look better.

## Calculate two rates with one clear denominator

The counts are invented to demonstrate the calculation.

| Metric | Formula | Illustrative result |
| --- | --- | --- |
| Mention rate | Answers naming the brand ÷ reviewed answers × 100 | 12 ÷ 40 × 100 = 30% |
| Site citation rate | Answers linking to the site ÷ reviewed answers × 100 | 5 ÷ 40 × 100 = 12.5% |

The two categories can overlap, so adding the percentages does not produce a total visibility rate. They do not need to be nested either: a link can appear without the brand name. Enter whole-number counts in the calculator, add the sample context and download the report. Keep the underlying responses with it; a percentage alone cannot explain why an answer was counted.

Always show counts beside percentages. Moving from one to two mentions in ten answers doubles the mention count, but it is still one additional observation. A small or selectively chosen prompt set can move sharply. Without a representative sampling design, do not attach a claim about the entire market to a precise-looking decimal.

## Compare equivalent samples and examine the misses

Repeat the same prompts under comparable conditions before interpreting a trend. If you add prompts, report the original set and the expanded set separately for the transition. Keep engine-specific reports rather than allowing a larger sample from one service to dominate a pooled average. Note product or model changes alongside changes you made to the website.

Then read the answers. An incorrect price, an outdated limitation and a missing citation suggest different follow-up work. Check whether your own public pages state the relevant fact clearly, whether supporting documentation exists and whether it is accessible. Improving a page can be worthwhile even if a later AI answer does not cite it. Avoid attributing every movement to the last edit.

## Keep business outcomes in a separate report

A mention rate is not click-through rate, and a citation rate is not a conversion rate. Where analytics can identify referred visits, review those visits and their outcomes separately. Google says traffic from its AI features is included in Search Console’s Web performance reporting; that does not turn a hand-collected answer sample into a dedicated AI traffic report. See [Google’s AI features documentation](https://developers.google.com/search/docs/appearance/ai-features) for its stated reporting scope.

> **A useful weekly review:** Save the sample definition, both counts, the raw answers, notable inaccuracies and the next action. This is enough to support a concrete discussion without a composite score that hides its inputs.

### Sources and useful next steps
- [AI visibility calculator](/tools/ai-visibility-calculator): Calculate and export the two sample rates.
- [Google: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features): Official eligibility and reporting guidance.
- [Check crawler access](/tools/ai-crawler-checker): Investigate a technical barrier separately from observed answer outcomes.

---

# Canonical URLs for launch campaigns, UTMs, and duplicate product pages

Canonical: https://rankvyze.com/blog/canonical-urls-for-launch-campaigns
Published: 2026-09-29
Category: Technical

Keep product launch tracking URLs, redirects, and duplicate pages from splitting search signals. A clear canonical workflow for founders.

A launch creates many URLs: a product page, announcement, directory listing, email link, social link, and UTM-tagged variations. Search engines need a clear primary URL for each piece of content. Analytics needs to know where visitors came from. Those goals can coexist.

## Choose the page that should appear in search

For a product page, the canonical is usually the clean, stable URL visitors would bookmark, such as /product. Campaign links can point to /product?utm_source=newsletter, while the page declares /product as its canonical. The tracking parameters may still be read by your analytics; the canonical communicates which version you prefer for indexing.

A directory profile hosted on another domain is a different page with its own context, reviews, or discussion. Do not assume it can or should canonicalize to your product site. Keep both pages useful and make their roles clear.

## Map common launch URLs

| URL type | Typical handling | Watch for |
| --- | --- | --- |
| Clean product page | Self-referencing canonical, linked in navigation | Accidental noindex or redirect |
| UTM-tagged link | Same content; canonical points to clean product URL | Parameter changing visible content |
| Old product slug | Permanent redirect to a close replacement | Redirecting unrelated pages to the homepage |
| Printable or duplicate view | Canonical to preferred equivalent when content is truly duplicate | Canonical to an unrelated page |
| Announcement article | Own canonical if it provides distinct news or analysis | Copying the product description verbatim |

## A five-minute verification

1. **Open both versions.** Visit the clean page and one tagged version. Confirm the content is equivalent and both URLs work.
2. **Read the HTML.** Inspect the rel=canonical value on each. It should be absolute or correctly resolved, stable, and match your chosen public URL.
3. **Follow internal links.** Navigation, sitemap, and related links should normally use the canonical URL. Campaign links may carry UTMs for measurement.
4. **Inspect after Google crawls.** Search Console URL Inspection distinguishes your declared canonical from Google's selected canonical. If they differ, investigate redirects, page similarity, and conflicting signals.

Canonical tags are hints, not commands. Google treats redirects as a strong canonical signal and sitemap inclusion as weaker evidence; conflicting signals make your preference less clear. A robots.txt block also prevents a crawler from reading the canonical on that page. Do not use a block as a shortcut for duplicate management.

> **Avoid a common launch mistake:** Do not canonicalize every new page to the homepage. If a launch announcement or comparison page has distinct, useful content, it deserves its own canonical URL.

For measurement, see [how to track product launch traffic with UTMs](/blog/track-product-launch-traffic-utm). For discovery, review [why a new site may not be indexed](/blog/why-your-new-site-isnt-indexed).

### Should UTM parameters be included in the sitemap?
Usually no. Include the preferred clean canonical URL in the sitemap; use tagged URLs in campaign links for attribution.

### Will a canonical tag fix a broken redirect?
No. Fix the redirect and destination directly. A canonical is a hint on a page that a crawler can reach and read.

### Sources and next steps
- [Google: How to specify a canonical URL](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls): Relative strengths of redirects, rel=canonical, sitemaps, and consistent links.
- [Google: Build and submit a sitemap](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap): Why sitemaps should contain preferred URLs.

---

# How to track product launch traffic with UTMs and activation events

Canonical: https://rankvyze.com/blog/track-product-launch-traffic-utm
Published: 2026-09-20
Category: Guides

Build a consistent UTM naming plan, verify campaign links, and connect launch visits to useful product actions without confusing attribution with causation.

A launch report should answer more than ‘How many visits did we get?’ It should show which placements brought visitors, what those visitors tried to do and where the product journey broke down. Campaign tags help with the first question. An explicit activation event helps with the second. Neither replaces reading the feedback people leave.

Use the [free UTM builder](/tools/utm-builder) to generate links from the naming plan below. The tool creates URLs locally; it does not install analytics or record visitors. You need a working analytics setup on the destination and a way to observe the product action you care about.

## Choose one campaign and a small naming vocabulary

Pick a campaign name that describes the initiative rather than the platform. For an illustrative autumn launch, use autumn-launch across the directory listing, announcement email and social post. Then vary source and medium to identify where each link is distributed. Use content when you need to distinguish placements or creative variants within a source.

Illustrative naming convention. Confirm how your analytics platform classifies each medium.

| Placement | utm_source | utm_medium | utm_campaign | utm_content |
| --- | --- | --- | --- | --- |
| RankVyze listing | rankvyze | referral | autumn-launch | product-profile |
| Your newsletter | product-newsletter | email | autumn-launch | main-button |
| An organic social post | linkedin | social | autumn-launch | demo-post |

Maintain this table in one shared place. Use consistent spelling and capitalization; Google’s [campaign URL guidance](https://support.google.com/analytics/answer/10917952?hl=en) explains that parameter values are case-sensitive. A source called RankVyze and another called rankvyze can fragment the report. Consistency is more useful than inventing a detailed naming taxonomy that nobody follows.

Keep personal data out of tags. An email address, customer name or private account identifier is not a campaign dimension. These values can appear in browser history, logs and downstream reporting. Describe the placement, not the individual clicking it. Use an approved internal identifier and an appropriate analytics design if you need account-level measurement elsewhere.

## Test the final URL, including redirects

Start with the real destination page and generate the tagged URL. If the destination already has a query parameter, the builder preserves it and correctly encodes the campaign values. It replaces the five UTM fields it controls. Empty optional content or term fields remove those existing tags, so review the output before publishing.

- Open the generated URL and confirm it lands on the intended page.
- Check whether a redirect drops the query string or changes the destination.
- Complete a test visit under the consent settings you intend to support.
- Verify the source, medium and campaign in your analytics debugging or reporting workflow.
- After publishing a listing, click the published link too; the platform may rewrite it.

Treat verification as part of launch preparation. A perfectly formatted URL is not proof that your analytics script loads, that consent permits collection or that a platform preserves parameters. If a directory does not permit tags, record the placement URL and publication time and examine available referral evidence. Do not manufacture a precise campaign attribution where the data does not support it.

## Define activation before you look at results

For a product that cleans CSV files, a completed export may be a more useful first action than opening the dashboard. For a planning tool, it might be creating and sharing a first plan. Choose the event that demonstrates a user reached initial value, document its definition and verify that it fires once under the intended conditions.

Keep the event stable during the measurement window. If you rename it, change when it fires or alter the product flow, annotate the report. Otherwise a tracking change can look like a change in customer interest. Use the same event definition across placements so a high-volume channel and a small niche community can be compared on the same basis.

| Metric | Question it answers | Important limit |
| --- | --- | --- |
| Attributed visits | Which tagged placements sent measurable sessions? | Some visits may be unobserved or lose their tags |
| Activated users | How many people reached the defined first value? | An event needs reliable implementation and deduplication |
| Activation rate | What share of the chosen visitor cohort activated? | Use compatible cohort and time-window definitions |
| Qualified enquiries | Did the channel reach people with a relevant need? | Qualification needs a consistent rule |

## Read the report at the right scope

In GA4, Google documents campaign dimensions in the Traffic acquisition report, including Session source/medium and Session campaign. Use session-scoped dimensions for a session acquisition question. Do not casually compare that total with first-user acquisition or a differently scoped conversion report and assume the rows represent the same population. Write the report name, date range and metric definition beside exported numbers.

If a link is copied into another community, its original tags may travel with it. If someone first discovers the product in a directory and later returns directly, the journey is more complex than one campaign click. Attribution is a reporting convention applied to observed data; it is not a controlled experiment showing what would have happened without the placement.

## Turn the review into an action list

For each channel, record one conclusion and one next step. If people arrive but fail at import, fix the importer before buying more exposure. If the audience asks the same question repeatedly, improve the listing and the relevant product page. If a placement produces no measurable traffic, first verify the link and tracking before judging audience fit. Save these notes with the campaign table for your next release.

### Sources and useful next steps
- [UTM campaign URL builder](/tools/utm-builder): Create and export correctly encoded links.
- [Google Analytics: custom campaign URLs](https://support.google.com/analytics/answer/10917952?hl=en): Official parameters and reporting guidance.
- [Product launch checklist](/tools/product-launch-checklist): Keep the measurement plan beside the rest of the launch tasks.

---

# How to rank on ChatGPT

Canonical: https://rankvyze.com/blog/how-to-rank-on-chatgpt
Published: 2026-09-03
Updated: 2026-09-04
Category: Guides

ChatGPT doesn't rank pages — it recommends businesses it understands. Here's how it actually picks, and the seven changes that get you named.

If you sell anything, some share of your buyers has already stopped typing keywords into Google and started asking an assistant a full question. **What's the best CRM for a two-person startup?** **Which Shopify agency should I hire for a fashion brand?** The answer that comes back names one or two businesses. Everyone else may as well not exist.

So the question every founder eventually asks is: how do I rank on ChatGPT? The honest answer starts by rejecting the premise.

## ChatGPT doesn't rank pages

Google returns an ordered list of documents. ChatGPT returns prose. There is no position four to climb to, no ranking factor to tune. What actually happens is closer to a recommendation than a ranking, and it comes from two places.

1. **What the model already knows.** Facts absorbed during training — who you are, what category you're in, who you serve. This is slow to change and you can't edit it directly.
2. **What it retrieves right now.** When a question needs current or specific information, ChatGPT searches the web, reads a handful of pages, and writes an answer citing them. This you *can* influence, and it's where nearly all the winnable ground is.

Both paths depend on the same thing: whether a machine reading your site can state plainly what your business is, who it's for, and why it's credible. Not whether your page ranks — whether your *entity* is legible.

> **Why good SEO isn't enough:** Ranking well means a page matched a query. Being recommended means a model understood a business well enough to vouch for it. Those are different jobs, and plenty of page-one sites fail the second one completely.

## The seven changes that actually move it

In roughly the order they pay off. The first three are where most businesses find their whole problem.

1. **1. Say what you are in the first sentence.** The single most common failure. A homepage H1 reading “We build beautiful things.” contains no noun a model can map to a category. Replace slogans with a plain definition: what you do, for whom, where. Put it in the H1 and the meta description, not below the fold.
2. **2. Make the claim machine-readable.** Add Organization schema with your name, URL, logo, areaServed and sameAs profiles, plus Service or Product schema on every offering page. This is the cheapest way to remove ambiguity, and it takes an afternoon.
3. **3. Render your content without JavaScript.** Most AI crawlers don't execute JS. If your homepage ships an empty div and fetches content client-side, your site is blank to them. Check by disabling JavaScript, or run curl and read what comes back.
4. **4. Publish the comparison content buyers ask for.** “Best X for Y”, “X vs Y”, “how much does X cost” dominate commercial intent. Engines answer them by citing pages that already frame the comparison. If only your competitor has written one, they own that answer.
5. **5. Earn corroboration off your own domain.** A claim on your site is one source. The same claim on directories, review sites, press and partner pages is evidence. Models weight agreement across domains heavily — it's how they avoid repeating marketing copy.
6. **6. Let the crawlers in, explicitly.** Check robots.txt for accidental blocks on GPTBot and OAI-SearchBot. Naming them in an Allow rule states intent unambiguously. Add /llms.txt describing your business and the pages you'd like cited.
7. **7. Structure answers as answers.** Question-shaped headings with direct answers underneath are the most quotable format there is. Lead each one with the sentence you'd want lifted verbatim — because that's exactly what happens.

## Two crawlers, two different jobs

A detail that trips people up: OpenAI runs more than one crawler, and blocking the wrong one has different consequences.

Current as of publication — check OpenAI's documentation for changes.

| User agent | What it does | Blocking it means |
| --- | --- | --- |
| GPTBot | Crawls for model training | You're less likely to be known by default |
| OAI-SearchBot | Indexes for ChatGPT search results | You can't appear in search-backed answers |
| ChatGPT-User | Fetches a page a user asked about | Users can't pull your page into a chat |

Blocking GPTBot is a legitimate choice if you don't want your content training a model. Blocking OAI-SearchBot while hoping to appear in ChatGPT's answers is simply a mistake, and a surprisingly common one.

## How to tell whether any of it worked

This is where most AEO advice stops, and it's the part that matters. You cannot improve what you don't measure, and there is no Search Console for ChatGPT.

What works is unglamorous: write down the 20 questions your buyers actually ask, ask each one in a fresh signed-out session, and record whether you were named, in what position, and whether any of your pages were cited. Repeat monthly. That's a real baseline, and it's the only way to know whether a change helped or you got lucky.

> **Use a signed-out session:** Logged in, ChatGPT personalises from your history and memory — including every previous time you asked about your own company. You'll see yourself mentioned and conclude you're visible. Always check the way a stranger would.

## How long it takes

Technical fixes — schema, rendering, crawler policy — can register within days to a few weeks, because retrieval reads your live site. Content and corroboration take longer, typically a month or two, because they depend on other pages being crawled and on the model finding agreement across sources. Anything baked into training weights moves on the model's release cycle, which you don't control.

In practice: expect the first movement in weeks, not days, and judge the work on whether mentions increase across several engines rather than on one lucky answer.

### Can you pay to rank on ChatGPT?
No. There is no paid placement in ChatGPT's organic answers. Anyone selling guaranteed ChatGPT rankings through payment to OpenAI is describing something that doesn't exist.

### Does traditional SEO help with ChatGPT?
It helps but isn't sufficient. Crawlability, clean structure and authority all carry over. What doesn't carry over is keyword-first thinking: ChatGPT needs to understand your business as an entity, not match a page to a phrase.

### How long does it take to show up in ChatGPT?
Technical changes like schema and crawler access can register within days to weeks. Content and third-party corroboration usually take one to two months. Training-derived knowledge changes only when the model is updated.

### Why does ChatGPT recommend my competitor instead of me?
Usually because it understands them better, not because it likes them more. They typically have a clearer entity definition, structured data, comparison content answering the exact question, and more third-party sources describing them the same way.

### Does blocking GPTBot hurt my visibility?
It reduces the chance the model learns about you during training, but it does not remove you from search-backed answers — that's OAI-SearchBot. Many sites block one while intending to block the other.

### Primary sources
- [OpenAI — GPTBot and crawler controls](https://platform.openai.com/docs/bots): The current user-agent strings and how to allow or block each one.
- [Schema.org](https://schema.org/docs/schemas.html): Canonical definitions for Organization, Service and FAQPage markup.
- [llmstxt.org](https://llmstxt.org): The /llms.txt convention referenced in step six.

---

# How to write a SaaS comparison page without inventing a winner

Canonical: https://rankvyze.com/blog/honest-saas-comparison-pages
Published: 2026-09-29
Category: Strategy

A comparison-page framework for SaaS founders that separates facts, opinion, missing data, and fit without fake reviews or misleading competitor claims.

A comparison page can answer a high-intent question: which product fits my situation? It can also become a thin sales page that makes unsupported claims about a competitor. The useful version shows its criteria, evidence, and limits so the reader can make the decision.

## Define the reader and the decision

Begin with a specific scenario, such as a two-person agency choosing a feedback tool, rather than a generic “Product A vs Product B” intro. State the date you reviewed the products and the jobs being compared. Different audiences may choose different winners.

## Use criteria that can be checked

| Criterion | Evidence to collect | How to present it |
| --- | --- | --- |
| Core workflow | Current product docs and hands-on use | Describe actual steps and limitations |
| Pricing | Official pricing pages and plan terms | State the checked date; link to current prices |
| Integrations | Official integration directories | Name native vs third-party connections |
| Data controls | Official security or help documentation | Avoid broad security claims you cannot verify |
| Onboarding | Actual trial flow or published setup docs | Explain required setup and time dependencies |
| Support | Published channels and response commitments | Do not invent service-level guarantees |

If you have not tested a feature, say “not verified” rather than guessing. If a capability is available only on a paid tier, identify the tier. Link to competitor documentation where it supports a factual statement, even though it may send a reader away. A comparison that survives scrutiny is more useful than one engineered only for a click.

## A structure that respects the reader

1. **Answer first.** Give a short decision summary by use case: who may prefer each option and why. Avoid declaring a universal winner.
2. **Show the evidence.** Use a concise criteria table with source links and the date checked. Separate product facts from your interpretation.
3. **Explain real tradeoffs.** For each major workflow, say what is easier, missing, or requires a workaround. Include a limitation of your own product.
4. **Offer next steps.** Link to both official product pages and a relevant trial or demo. Make a reader's next action clear.
5. **Schedule maintenance.** Recheck pricing and major feature claims at a set interval or when either product announces changes.

## Search and AI-search considerations

Google's people-first guidance asks whether a page adds original value beyond what others publish. A template filled with swapped competitor names and identical text will rarely do that. Firsthand notes, transparent sources, specific use cases, and candid limits create a better answer. Search engines may still choose a different page; no comparison format guarantees ranking or AI citation.

Keep comparisons as distinct pages only when you can provide distinct evidence. Use a descriptive title and link from related product or resource pages. If two pages answer essentially the same question, consolidate them instead of multiplying near-duplicates.

> **Editorial disclosure:** If you sell one of the products being compared, say so near the beginning. The reader should understand your relationship before weighing your recommendation.

For the surrounding product page, use the [product page guide](/blog/product-pages-for-search-and-ai). For internal link planning, use the [internal linking guide](/blog/internal-linking-for-ai-search).

### Should we publish a page for every competitor?
Only when each page answers a real comparison question with current, specific evidence. Generating many near-identical pages for keyword variants can be low-value and may conflict with search spam policies.

### Can we quote a competitor's price?
You can report a price accurately with a source and checked date, but pricing changes. Link to the official pricing page and update your comparison promptly.

### Sources and next steps
- [Google: Creating helpful, reliable, people-first content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content): Quality questions for original comparison content.
- [Google Search spam policies](https://developers.google.com/search/docs/essentials/spam-policies): Scaled-content and misleading-content considerations.

---

# Build product pages that answer real buyer questions in search and AI

Canonical: https://rankvyze.com/blog/product-pages-for-search-and-ai
Published: 2026-09-20
Category: Strategy

Turn a vague product landing page into a useful decision page with clear facts, visible limitations, worked examples and supporting documentation.

A buyer evaluating software needs to know what it does, whether it fits their workflow and what it will cost to use. Many product pages make those questions surprisingly difficult to answer. They lead with abstract benefits, show an attractive interface and leave practical limits to a sales conversation or a support thread.

A more useful page makes important facts easy to find and easy to verify. That is a sound content goal for people arriving from a directory, conventional search or an AI answer. It does not require a claim that a particular page format will force a citation. The structure below is an editorial template, not a discovered ranking formula.

## Write the product identity in one concrete paragraph

Use a simple starting pattern: product name, intended user, primary task and meaningful constraint. An illustrative example is: ‘TableClean helps operations teams standardize CSV exports before importing them into a CRM. It detects duplicate rows and inconsistent date formats. The browser demo supports files up to the limit shown below.’ The example product is fictional; the point is the specificity of the explanation.

Compare that with ‘Unlock the future of intelligent data workflows.’ The second sentence leaves the reader to infer the category and capability. You can still express a distinctive brand voice, but give visitors enough concrete information to decide whether to continue. Keep names and product descriptions consistent across the homepage, launch profile, documentation and pricing page.

## Build a visible fact table before adding promotional claims

| Buyer question | What the page should state | Supporting evidence |
| --- | --- | --- |
| Will it work with my data? | Supported formats, integrations and prerequisites | A setup guide or reproducible example |
| What is included? | Plan features and meaningful usage limits | The current pricing page |
| What happens to my data? | The actual handling and retention behavior | Applicable product documentation and policy |
| Who is it unsuitable for? | Known exclusions and unsupported workflows | A limitation section or compatibility guide |
| How do I start? | A specific first action and expected output | A working demo or onboarding path |

Only publish facts you can maintain. If an integration is planned, label it as planned instead of listing it as available. If a capability requires a paid tier or a separate service, say so beside the claim. Avoid filling gaps with optimistic assumptions. A clear limitation can save an unsuitable customer time and make a suitable customer’s evaluation easier.

## Show one complete workflow

Choose an example that resembles the task a buyer is trying to complete. Explain the starting input, the action taken and the resulting output. For the fictional CSV product, show a small input with two date formats, the normalization setting and the output. State which changes were automatic and which required a choice. Include a failure case if it helps someone understand the boundary of the feature.

Use screenshots to support the explanation rather than replace it. Keep captions readable, give meaningful images useful alternative text and provide the key facts as visible text. Link to detailed documentation where needed. A buyer should be able to understand the workflow even if they cannot view the screenshot or do not want to play a video.

## Connect product claims to evidence

Create a claim ledger before publishing. For each strong statement, record where the evidence lives, who owns it and when it needs review. A feature claim may point to a testable demo; an integration claim to a setup guide; a performance claim to a methodology with conditions. If there is no defensible evidence for ‘fastest’ or ‘most accurate,’ replace the superlative with a description of what the product does.

| Claim | Useful qualification |
| --- | --- |
| Exports to CSV | Specify which fields are included and on which plan |
| Works without an account | Explain which demo actions are available before signup |
| Processes a file quickly | Describe the tested file size, environment and measurement if reporting a time |
| Integrates with a CRM | Name the supported integration method and prerequisites |

Keep this evidence near the relevant statement. A general research link in the footer does not substantiate every claim on a feature page. If you publish a testimonial or case study, use a real, authorized account of the result with its context. Do not invent a customer to make the page look established.

## Apply search fundamentals without a special AI layer

Google’s [AI features guidance](https://developers.google.com/search/docs/appearance/ai-features) says there is no special AI markup or additional technical requirement for its AI search features. Pages need to be eligible in Search, and inclusion is not guaranteed. Its practical guidance includes accessible text, internal discovery links and structured data consistent with the visible page. These are foundations to check, not promises of recommendation.

Inspect your live page with [What AI Crawlers See](/tools/what-ai-crawlers-see) and the [Schema Checker](/tools/schema-markup-checker). Treat tool output as diagnostic evidence and investigate any mismatch. If a price, capability or organization fact exists only in structured data but not on the page, fix the disagreement instead of adding more markup.

## Maintain the page as the product changes

Assign an owner to review the page when pricing, integrations or onboarding changes. Keep the launch listing aligned with the same factual source. Use customer questions to identify missing explanations, and record important edits so later reports have context. The useful outcome is a page that answers the decision clearly—not a longer page built around repeated keywords.

### Sources and useful next steps
- [Google: AI features](https://developers.google.com/search/docs/appearance/ai-features): Official eligibility and content guidance.
- [Inspect crawler-visible content](/tools/what-ai-crawlers-see): Check what a fetched page exposes.
- [Plan a launch](/blog/product-launch-seo-checklist): Connect the product page to distribution and follow-up.

---

# Write SaaS title tags and meta descriptions that make the page clear

Canonical: https://rankvyze.com/blog/saas-title-tags-meta-descriptions
Published: 2026-09-20
Category: Technical

Draft distinct titles and descriptions for product, feature, integration and comparison pages, then verify the deployed tags and their fit with the visible content.

A useful search title tells someone what page they are about to open. A useful description gives them a reason to believe it answers their question. For a SaaS site, that means distinguishing the homepage from a feature page, a setup guide, a pricing page and a comparison. Repeating the same broad promise across all of them makes the choice harder.

Use the [SERP Snippet Preview](/tools/serp-snippet-preview) to compare drafts and export title and description tags. The preview is an editing aid. It does not fetch Google results, predict click-through rate or guarantee that the search engine will show the text exactly as entered.

## Start with the page’s job

Write a one-sentence brief before drafting a title: ‘This page helps this reader do this task.’ A pricing page helps someone understand plans and limits. An integration page helps them assess compatibility or complete setup. A comparison page helps them understand tradeoffs between defined options. If two pages have the same brief, decide whether they need clearer roles before trying to distinguish them with synonyms.

TableClean is a fictional example. Drafts describe page roles, not tested ranking results.

| Page type | Weak draft | More specific illustrative draft |
| --- | --- | --- |
| Product homepage | Welcome to TableClean | CSV Cleanup for Operations Teams | TableClean |
| Feature page | Powerful Features | TableClean | Find Duplicate CSV Rows | TableClean |
| Integration guide | Connect Everything | Prepare CSV Files for CRM Import | TableClean |
| Pricing page | Get Started Today | TableClean Pricing: Plans and File Limits |

The improvement is not that every title follows a magic template. It is that the subject becomes clear. Put the most distinguishing information where a reader can see it quickly, and use brand text consistently without allowing it to crowd out the page’s task. A narrowly useful page does not need to pretend to be the entire product.

## Make the description a truthful preview

Summarize what the visitor will find and include a meaningful detail when it helps the decision. For the fictional duplicate-row feature, a draft could say: ‘Find repeated rows in a CSV, choose the columns used for matching and preview changes before exporting. See supported file limits and a worked example.’ Every capability mentioned should actually be explained on the destination.

Avoid describing a paid capability as free, implying an integration exists when it is only planned or promising an outcome you cannot support. A compelling description that sends someone to the wrong experience is not a good result. Read the title, description, H1 and opening paragraph together; they should describe the same page from slightly different angles.

## Use counts for editing, not as ranking rules

Google’s [title-link documentation](https://developers.google.com/search/docs/appearance/title-link) recommends descriptive, concise titles and explains that displayed titles can be generated from several sources. It does not prescribe a universal character limit that guarantees display. Device width and the text itself affect truncation. Use a preview to notice a buried meaning, rather than trimming a useful word solely to hit a number.

Google also explains that [snippets are generated primarily from page content](https://developers.google.com/search/docs/appearance/snippet), with the meta description used when appropriate. Different queries can produce different snippets. Your task is to supply a good summary and useful page content; you do not control every version shown in search.

Try the mobile view in the preview tool and ask whether the visible beginning still makes sense. Replace redundant phrasing before removing the detail that distinguishes the page. Character counters are useful for comparing two drafts, but a green counter would not prove that a title is accurate, readable or likely to be used.

## Publish through one source of truth

If your CMS has SEO fields, use them instead of pasting additional tags into the body. If you manage the document head in code, make sure layout defaults and page-specific metadata produce one intended title and description. The preview tool escapes special characters in exported HTML, but your framework may expect plain strings in its metadata API rather than raw tag markup.

- Apply the draft to the correct route, not only a local preview.
- Inspect the deployed page with the Meta Tag Checker and confirm the actual output.
- Check that the canonical points to the intended public destination.
- Check whether the framework appends the brand name a second time.
- Verify that the main heading and visible content still support the search copy.

This last step catches a common handoff problem: a writer supplies a complete title including the brand while the website already adds the brand automatically. The result can be repetitive and longer than intended. Define whether editorial drafts include the suffix, and keep that convention consistent across the site.

## Review changes with context

Record the old text, new text and deployment date. When you review search performance later, inspect relevant pages and queries rather than attributing a site-wide movement to one title edit. Seasonality, position changes, query mix and other site edits can affect clicks. A before-and-after comparison is useful evidence, but it is not automatically a controlled test of the wording.

If Google displays a different title, inspect whether your draft matches the page, whether prominent headings disagree and whether old wording remains in important places. Do not keep rewriting a useful title simply because one query produced a different result. Focus on accurate page identity and the experience after the click.

### Does this preview show my current Google result?
No. It renders text you enter in an illustrative layout. Use it for drafting, and use the Meta Tag Checker to inspect tags served by your website.

### Should every product page use the same description?
Distinct pages should explain their own purpose. Reuse brand facts where appropriate, but describe the task and information available at that destination.

### Sources and useful next steps
- [SERP snippet preview](/tools/serp-snippet-preview): Draft, compare and export search copy.
- [Meta Tag Checker](/tools/meta-tag-checker): Inspect the tags served by the deployed page.
- [Google: title links](https://developers.google.com/search/docs/appearance/title-link): How titles are generated and common problems.
- [Google: snippets](https://developers.google.com/search/docs/appearance/snippet): How descriptions and page content inform snippets.

---

# AEO tools vs AEO services: which one do you actually need?

Canonical: https://rankvyze.com/blog/aeo-tools-vs-aeo-services
Published: 2026-09-07
Category: Strategy

An AEO tool tells you whether AI engines mention you. A service changes whether they do. Real prices, and how to tell which one you actually need.

**An AEO tool tells you whether AI engines mention you. An AEO service changes whether they do.** They are sold in the same aisle, priced within range of each other, and solve entirely different problems — and picking the wrong one is the most common way money gets wasted in this category.

> **Where this comes from:** We read the roundups that rank for "best AI SEO tools", "best GEO tools" and "best AI visibility tools". Every one of them reviews software, and only software. Not one includes a done-for-you service. If you searched those terms hoping to find someone to do the work, the results were never going to contain the answer.

## The difference in one table

|  | AEO tool | AEO service |
| --- | --- | --- |
| What you buy | A dashboard and a data feed | The work, done |
| What it produces | Reports: which prompts mention you, which cite competitors | Changes to your site and your entity footprint |
| Typical price | $50–$500/month, ongoing | $99 one-time to $10,000/month |
| Who operates it | You, or your team | The provider |
| Fails when | Nobody acts on the report | You cannot tell what was actually done |
| Right for | Teams with a developer and a writer already | Teams without one, or without the hours |

The trap is that a tool feels like progress. You buy it, the dashboard fills with prompts, and you can see precisely how invisible you are — in colour, updated weekly. None of that changes an answer. Measurement is not the intervention.

## What a tool genuinely gives you

This is not an argument against tools. Tracking is real work that is tedious to do by hand, and the good ones do it properly: running your buyer's questions across engines on a schedule, logging which brands appear, capturing the URLs each answer cites. That last part is the underrated one — citation sources tell you which pages an engine already trusts on your topic, which is a list of where to get mentioned.

A tool earns its subscription when someone on your team will read the output and act within the week. If that person does not exist, you are paying a monthly fee to be told bad news.

## What a service gives you, and what to check

A service should be judged on what changed, not on what was reported. The work in AEO is unglamorous and specific: making sure your content is in the HTML rather than assembled by JavaScript a crawler will not run, stating who you are in structured data, getting the AI search crawlers unblocked, and building the third-party corroboration engines cross-check you against.

- **Ask what gets changed.** A provider who cannot name the specific edits is selling a report with a bigger invoice.
- **Ask about crawler access first.** If AI search crawlers cannot fetch your pages, everything else is decoration. Our [AI crawler checker](/tools/ai-crawler-checker) answers this in a few seconds, free.
- **Ask how they measure.** "Mentioned on two engines" is checkable. "Improved AI visibility" is not.
- **Ask what happens if it does not work.** A provider carrying no risk has no reason to prioritise you.

## How to tell which one you need

1. **Check whether you are retrievable at all.** Run the [AI crawler checker](/tools/ai-crawler-checker) and [what AI crawlers see](/tools/what-ai-crawlers-see). If crawlers are blocked, or your content only exists after JavaScript runs, no tool subscription will help — this is an implementation problem.
2. **Decide who will act on a report.** Name the person. If you cannot, a tool will produce dashboards nobody opens. Buy the work instead.
3. **Check whether anything about you exists off-site.** Search your brand name in quotes. If no third party mentions you anywhere, tracking will confirm that for months while nothing changes. Corroboration is the constraint, and it is service work.
4. **Buy tracking once there is something to track.** Once you are retrievable, identifiable and mentioned somewhere, a tool starts earning its fee — it tells you which of those mentions actually moved an answer.

## Where we sit, plainly

RankVyze is a service, not a tool. We charge $99 once for a 45-day sprint, and if you are not mentioned on at least two AI engines by the end of it, we refund the whole thing. We publish that number because almost nobody in this category publishes any number, which makes it impossible to budget for the work.

If what you actually need is tracking, buy tracking — we compare the real options and their real prices on [AI visibility tools](/ai-visibility-tools), including the ones we do not compete with.

### Can I do AEO myself?
Yes, and if you have a developer and a writer it is often the right call. The work is roughly 20 to 40 hours done properly: server-side rendering for content that matters, structured data, crawler policy, and off-site profiles that agree with each other. Our free tools cover the diagnostic half at no cost.

### Do AEO tools actually work?
For measurement, yes — the established ones run real prompts across real engines and report what came back. What none of them do is change the answer. Treat them as instrumentation, and budget separately for whatever the instrumentation tells you to fix.

### Why do the 'best AI SEO tools' lists never include services?
Because the query asks for software, and the publishers are usually software vendors or agencies reviewing software. It is not a conspiracy, it is category convention — but it does mean the lists are the wrong place to look if you want the work done rather than measured.

### How long before an AI engine mentions a business?
For a site that is already indexed and has some third-party presence, typically four to eight weeks after the fixes land. For a brand-new domain with no mentions anywhere, longer — the constraint is corroboration from other sites, and that cannot be rushed by anything on your own server.

### Related
- [What AEO actually costs](/blog/what-answer-engine-optimization-costs): The four ways to buy it, with real ranges.
- [AI visibility tools compared](/ai-visibility-tools): Real prices, read from the vendors' own pages.
- [How to choose an AEO tool](/blog/how-to-choose-an-aeo-tool): If tracking is what you need.

---

# What answer engine optimization actually costs

Canonical: https://rankvyze.com/blog/what-answer-engine-optimization-costs
Published: 2026-09-04
Category: Strategy

AEO pricing ranges from $0 to $10,000 a month. The real ranges for tools, retainers and fixed-scope sprints, what drives the number, and how to tell which you need.

**Answer engine optimization costs between $0 and about $10,000 a month, and the spread is almost entirely about who does the work rather than how much gets done.** Here are the actual ranges, what sits inside each, and how to tell which one you need.

> **Why this page exists:** Nearly every AEO provider hides pricing behind a call. That is a choice about sales process, not a fact about the work — and it makes the category impossible to budget for. The numbers below are what we see quoted; ours is at the bottom, stated plainly.

## The four ways to buy it

| Option | Typical cost | What you get | Best when |
| --- | --- | --- | --- |
| Do it yourself | $0 plus your time | Schema, rendering fixes, crawler policy, content — roughly 20–40 hours to do properly | You have a developer and a writer already |
| AEO tool subscription | $50–$500/month | Tracking and reporting. Tells you where you stand; does not fix anything | You have a team who will act on the data |
| Retainer agency | $2,000–$10,000/month | Ongoing research, audits, implementation, reporting | You need continuous work across many pages |
| Fixed-scope sprint | $99–$5,000 one-time | A defined engagement: measure, fix, re-measure | You want to know whether this works before committing |

The gap between a tool subscription and an agency retainer is where most people get stuck. A tool tells you that ChatGPT doesn't mention you. It cannot tell you that your homepage H1 says nothing a machine can categorise, and it certainly can't rewrite it.

## What actually drives the price

- **Whether implementation is included.** Diagnosis is a few hours. Doing the work — schema, rendering, entity copy, new pages — is most of the cost. Read any quote for this first.
- **How research is done.** Manual analysts asking real questions on real engines cost more per check than an API, and produce results that reflect what a customer actually sees. Automated checks are cheaper and drift from reality.
- **How many questions are tracked.** Ten prompts across four engines is 40 checks per round. A hundred prompts is 400. This scales linearly and is the single biggest input to an agency's number.
- **Whether anything is guaranteed.** Almost nothing in this category is. A provider carrying the risk has priced it in — or has decided it can meet the bar.

> **The question that filters most providers:** Ask: “What exactly do I get if this doesn't work?” The common answer is a report explaining why the market was difficult. If there is no defined outcome and no consequence for missing it, you are buying activity, not a result.

## Is it worth paying for at all?

Only if your buyers ask comparative questions before choosing. If someone already knows your brand name and types it, an AI answer isn't in the path. If they ask “what's the best X for Y”, it is — and the answer names two or three companies.

1. **Work out what one customer is worth.** Average deal value times gross margin. For most B2B services this is in the thousands.
2. **Count the questions you'd want to win.** Ten to twenty is typical for a focused business.
3. **Check where you stand on them today.** Ask each one in a signed-out session across ChatGPT, Perplexity, Gemini and Claude. Count how often you're named.
4. **Compare that gap to the price.** If you are absent from questions your buyers ask and one customer covers the cost several times over, the arithmetic is not close.

## What RankVyze costs

**$99, once.** That covers a 45-day sprint: baseline research across ChatGPT, Perplexity, Gemini and Claude, a full AEO audit scored across six categories, implementation of the fixes as reviewable changes, and a re-measurement against the day-zero baseline.

If your business isn't mentioned on at least two of the four engines by the end of those 45 days, on the prompt set agreed at the start, the $99 is refunded in full. The conditions that void that are [published in advance](/guarantee) — there are no others.

It is a fixed-scope sprint, not a retainer, and it is deliberately priced to remove the decision. See [what's included](/pricing), or read about [the service itself](/answer-engine-optimization).

### How much does answer engine optimization cost?
Anywhere from $0 doing it yourself to $10,000 a month on an agency retainer. Tool subscriptions run $50–$500 a month but only measure. Fixed-scope sprints run $99–$5,000 one-time. RankVyze is $99 once for a 45-day sprint, refunded if it doesn't work.

### Why do most AEO agencies not publish prices?
Because scope varies and because a call converts better than a price page. It isn't necessarily a red flag, but it does mean you cannot compare providers without several conversations — which is itself a cost.

### Is a monthly AEO retainer worth it?
It can be, if you have many pages, many questions to win, and a competitor actively working the same ground. For a business with ten to twenty buyer questions and one site, a fixed-scope engagement usually reaches the same place for far less.

### Can I do answer engine optimization myself for free?
Yes. The technical half — Organization and Service schema, server-rendered content, crawler access, a plain entity definition on the homepage — is roughly a day's work for a competent developer. The slower half is content and third-party corroboration.

### Does AEO pricing include content writing?
Often not, and it is the most common hidden cost. Comparison and pricing pages are usually what wins commercial questions, so if writing is excluded you are buying diagnosis rather than a result. Ask explicitly.

---

# Internal linking for AI search: the pages your own site votes for

Canonical: https://rankvyze.com/blog/internal-linking-for-ai-search
Published: 2026-09-07
Category: Technical

Why sitewide links tell you nothing, what an orphan page really costs, and how to find both — with real numbers from crawling our own site.

**Internal links are the only ranking signal you control completely.** No outreach, no waiting, no one else's approval. And on most sites they are quietly broken in a way nobody notices, because the thing that breaks is invisible in every dashboard: the pages your own site says are important are not the pages you would name.

## Why this matters more for AI search than it used to

Engines use internal links for two things: finding pages, and judging which ones matter. For AI answers the first one dominates. An answer engine has to retrieve a page before it can quote it, and retrieval depends on the page having been crawled recently enough to be in the index at all. A page nothing links to gets crawled once, if the sitemap is read, and then largely forgotten.

That is a harsher outcome than it used to be. In traditional search a rarely-crawled page still sits in the index and can surface for a long-tail query. In AI search, a page that was not retrieved simply does not exist for that answer.

## The measurement mistake almost everyone makes

Count inbound internal links naively and your privacy policy wins. It is in the footer of every page, so on a 200-page site it collects 200 inbound links — more than any article you have ever written. Your terms page comes second. This is not a quirk to work around; it is the reason most internal link reports are useless.

> **The distinction that makes the number mean something:** Separate sitewide links from contextual ones. A sitewide link appears on nearly every page, so it cannot distinguish an important page from an unimportant one. A contextual link is one a person chose to place in a sentence. Only the second kind tells you anything, and it is the kind engines weigh most heavily.

Our [internal link checker](/tools/internal-link-checker) does this split automatically: any target appearing on 80% or more of the crawled pages is classed as template furniture and reported separately, leaving a ranking of pages that earned their links.

## Three problems worth finding

### Orphan pages

A page nothing links to. It is usually in the sitemap, so it can be discovered — but discovery and inclusion are separate decisions, and a page with no inbound links gives an engine no evidence that it matters. Orphans are the single most common reason a page stays unindexed on an otherwise healthy site, and the fix is almost never technical. It is one contextual link from a page that already gets crawled.

### Click depth

Crawl frequency falls off sharply with distance from your homepage. Two clicks gets revisited regularly. Four clicks might be fetched once and left for months. That matters most for anything whose value depends on being current — pricing, availability, comparison tables with dates in them.

### Broken internal links

Every one spends crawl budget on nothing and lands a reader on an error. They accumulate silently after any restructure, because nothing warns you when a link you wrote two years ago stops resolving.

## What our own site looked like

We ran the checker against rankvyze.com while building it, which is the only honest way to describe what the output is worth. At the time of writing the site had 93 URLs in its sitemap. A 25-page crawl found 670 internal links across 94 unique targets.

| Clicks from homepage | Pages | What it means |
| --- | --- | --- |
| 0 | 1 | The homepage itself |
| 1 | 24 | Everything in the nav and footer |
| 2 | 69 | The long tail — glossary, guides, tools |
| 3 or more | 0 | Nothing buried |

Zero orphans and zero broken links, which is the boring answer you want. But the finding that mattered was in the ranking rather than the errors: our highest-intent commercial page was reachable only through one hub page, two clicks from anywhere, while eight lower-intent pages sat in the footer. Nothing was broken. The site was simply voting for the wrong things.

## How to fix it, in order

1. **Find the orphans.** Run the [internal link checker](/tools/internal-link-checker) against your homepage. Anything in your sitemap that nothing links to is the first list to work through, because those pages are getting no consideration at all.
2. **Give each orphan one real link.** From a page that is already crawled well, inside a sentence, with anchor text that describes the destination. One contextual link beats being added to the footer.
3. **Pull your commercial pages up to two clicks.** List the pages that make money. If any is three or more clicks from the homepage, it needs a link from somewhere shallower — usually the nav, the footer, or a hub page that is already shallow.
4. **Fix broken links and stray nofollows.** Both are cheap to fix and pure loss otherwise. Internal nofollow is almost always unintentional and usually arrives as a plugin default.
5. **Re-run and compare.** The ranking should now put your commercially important pages near the top. If it does not, your site is still telling engines something different from what you intend.

### How many internal links should a page have?
There is no correct number, and targets like "100 links per page" are folklore. What matters is that every page you care about receives at least one contextual link, and that important pages receive more than unimportant ones. Ranking your pages by contextual inbound links and checking the order matches your priorities is a better test than any count.

### Do footer links count?
They count for discovery — a footer link guarantees a page is reachable and crawled. They count for very little in judging importance, because every page has them. Use the footer to guarantee reachability, and body links to signal importance.

### What is a good click depth?
Within two clicks of the homepage for anything commercially important. Three is acceptable for archive and reference content. Beyond three, expect infrequent crawling and slow updates.

### Does anchor text matter for AI search?
Yes, and arguably more than for traditional search. Anchor text is one of the clearest statements on your site about what a page is, and engines building an understanding of your entity read it as a label. "Click here" wastes that; the destination's actual subject does not.

### Can internal linking fix a page that is not indexed?
Often, yes — if the reason it is unindexed is that nothing links to it, which is common. It will not fix a page blocked by robots.txt, carrying a noindex tag, or whose content only exists after JavaScript runs. Rule those out first with the [meta tag checker](/tools/meta-tag-checker).

### Tools used in this post
- [Internal link checker](/tools/internal-link-checker): Ranks your pages by contextual links, finds orphans and depth problems.
- [Sitemap checker](/tools/sitemap-checker): Confirms what you are actually declaring to crawlers.
- [Meta tag checker](/tools/meta-tag-checker): Rules out noindex and canonical problems.

---

# Why your business doesn't show up in ChatGPT

Canonical: https://rankvyze.com/blog/why-your-business-doesnt-show-up-in-chatgpt
Published: 2026-09-03
Category: Guides

Six reasons a real business stays invisible in AI answers, in the order they're worth checking — and how to tell which one is yours in about twenty minutes.

You ask ChatGPT the question your best customers ask. It names three companies. None of them is you — and at least one is objectively worse at the job. This is the moment most people go looking for an answer, and the useful thing to know is that invisibility is rarely mysterious. It has a cause, and the cause is usually findable in under half an hour.

Here are the six causes in the order they're worth checking, cheapest first.

## 1. The crawlers can't read your site

Start here because it's two minutes and it invalidates everything else. Most AI crawlers do not execute JavaScript. If your content is fetched client-side, the page they see is empty.

If that prints almost nothing, you've found your problem.

```bash
# What a non-JS crawler actually sees
curl -s https://yoursite.com | sed 's/<[^>]*>//g' | tr -s '[:space:]' ' ' | head -c 600
```

Then check `robots.txt` for rules blocking `GPTBot` or `OAI-SearchBot`. Plenty of sites blocked AI crawlers in 2023 on general principle and never revisited the decision.

## 2. Your homepage doesn't say what you are

The most common cause among sites that are otherwise healthy. Marketing copy optimises for feeling; models need category. An H1 reading “Growth, unlocked.” tells a model nothing it can match to “best analytics tool for ecommerce”.

The test: read your homepage's first hundred words and ask whether a stranger could finish the sentence *“This company is a ___ that helps ___ do ___.”* If they can't, neither can a model.

| Instead of | Write |
| --- | --- |
| We build beautiful things. | Shopify agency for fashion and lifestyle brands. |
| Growth, unlocked. | Analytics for ecommerce teams who want to understand checkout drop-off. |
| Your partner in success. | Immigration law firm for startup founders relocating to Canada. |

## 3. Nothing on your site is machine-readable

Without structured data, a model has to infer your identity from prose. With it, you're simply telling it. Organization schema is the floor; Service or Product schema on offering pages is what answers “best X for Y” questions.

The minimum viable Organization block. sameAs is doing more work than it looks.

```json
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Acme",
  "url": "https://acme.com",
  "description": "Shopify agency for fashion and lifestyle brands.",
  "areaServed": ["US", "GB", "IN"],
  "sameAs": [
    "https://www.linkedin.com/company/acme",
    "https://www.shopify.com/partners/acme"
  ]
}
```

## 4. You've never answered the question being asked

Retrieval-backed answers cite pages that already address the query. If the question is “how much does a Shopify agency cost” and your pricing page redirects to a contact form, there is nothing to cite. A competitor who published honest ranges gets the citation by default.

List the ten questions a buyer asks before choosing you. Count how many have a page that answers them directly. For most businesses the answer is two or three.

## 5. Only you say what you are

Models weight corroboration because it's the cheapest defence against believing marketing copy. If acme.com is the sole source describing Acme as a fashion Shopify specialist, that claim is weak. If a directory, two review sites and a press mention describe it the same way, it's a fact.

> **The cheapest corroboration you already have:** Most businesses are already listed somewhere they've forgotten about — a partner directory, an old press release, a marketplace profile. Find them, make sure they describe you consistently, and link them from your Organization schema's sameAs array.

## 6. You're actually visible and testing it wrong

Worth ruling out before spending money. Two mistakes account for nearly all false readings.

- **Testing while signed in.** ChatGPT personalises from your history and memory. If you've discussed your company before, it will happily mention it — to you, and to nobody else.
- **Testing one prompt once.** Answers vary between sessions. A single miss means little; a miss across ten prompts and four engines is a finding.

## A twenty-minute diagnosis

1. **Minutes 0–2.** curl your homepage. Confirm real text comes back without JavaScript.
2. **Minutes 2–4.** Open /robots.txt. Confirm nothing blocks GPTBot or OAI-SearchBot.
3. **Minutes 4–6.** View source, search for ld+json. Note whether Organization exists at all.
4. **Minutes 6–10.** Read your first hundred words. Complete the “a ___ that helps ___” sentence.
5. **Minutes 10–20.** In a signed-out session, ask five real buyer questions across ChatGPT and Perplexity. Record who gets named and which pages get cited.

At the end you'll have a specific cause rather than a vague worry — and the fix for each of these is a known quantity, not a mystery. If you'd rather not run it yourself, that's roughly what our [free scan](/pricing) automates for the technical half.

### Why does ChatGPT know my competitor but not me?
Almost always because their site states its category plainly, carries structured data, answers the specific question being asked, and is described consistently on other domains. It's legibility, not preference.

### Does ChatGPT have an index I can submit to?
There is no submission form equivalent to Google Search Console. Search-backed answers rely on crawling, so the controls you have are robots.txt, your site's structure, and being cited elsewhere.

### How do I check if ChatGPT can see my website?
Fetch your homepage without JavaScript and read what comes back, then confirm robots.txt doesn't block OAI-SearchBot. If the raw HTML has little text, crawlers see little text.

### Check these directly
- [OpenAI — GPTBot and crawler controls](https://platform.openai.com/docs/bots): Which user agent does what, and how to allow each.
- [Google — crawler and fetcher overview](https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers): Includes Google-Extended, the control for Gemini.

---

# Why your new site isn't indexed yet

Canonical: https://rankvyze.com/blog/why-your-new-site-isnt-indexed
Published: 2026-09-07
Category: Technical

Five things that genuinely block indexing, the submission protocol most people have never heard of, and a realistic timeline for a new domain.

**A new domain typically takes one to four weeks for its first pages to appear in Google, and longer to be indexed in full.** Before assuming something is broken, it is worth ruling out the handful of things that genuinely block indexing — and then using the one submission route most people have never heard of.

> **First, check your search syntax:** The site: operator takes no space. Searching site: yoursite.com with a space silently disables the operator and returns ordinary results, including other companies with similar names. It is site:yoursite.com, and getting this wrong has convinced a lot of people they had a problem they did not have.

## Five things that actually block indexing

Work through these before concluding it is a waiting game. Each takes under a minute and each is a real, common cause.

| Check | What kills you | How to check |
| --- | --- | --- |
| robots.txt | Disallow: / left over from staging | Load yoursite.com/robots.txt and read it |
| noindex | A meta robots noindex shipped to production | [Meta tag checker](/tools/meta-tag-checker) |
| X-Robots-Tag | A noindex header set at the CDN, invisible in the HTML | Check response headers |
| Canonical | Every page pointing at the homepage | [Meta tag checker](/tools/meta-tag-checker) |
| Rendering | Content assembled by JavaScript the crawler never runs | [What AI crawlers see](/tools/what-ai-crawlers-see) |

A staging-era robots.txt is the most common single cause, and the most embarrassing, because the site works perfectly for every human who visits it. The noindex header is the nastiest, because it does not appear anywhere in your HTML — you have to look at the response headers to find it.

## The submission route nobody mentions

Google gets all the attention, and Google has exactly one submission mechanism: Search Console. Verify the property, submit the sitemap, request indexing on individual URLs. There is no faster path and no API that changes this.

Bing is different, and this is the part worth knowing. Bing supports **IndexNow**, an open protocol that lets you push URLs directly rather than wait to be crawled. You host a key file at your domain root, POST a JSON body listing your URLs, and Bing, Yandex and Seznam all receive the same submission.

POST to https://api.indexnow.org/indexnow

```json
{
  "host": "yoursite.com",
  "key": "your-key",
  "keyLocation": "https://yoursite.com/your-key.txt",
  "urlList": [
    "https://yoursite.com/",
    "https://yoursite.com/pricing"
  ]
}
```

> **Why Bing matters more than its market share suggests:** ChatGPT's search step is Bing-backed. A page Bing has not indexed cannot be retrieved, summarised or cited in a ChatGPT answer — so on a new domain, Bing indexing latency is the actual bottleneck on AI visibility, and it is the one you can do something about today.

Up to 10,000 URLs go in a single request. Do not resubmit unchanged URLs repeatedly — the endpoint returns a 422 for that, and it is treated as spam behaviour.

## What actually speeds Google up

1. **Verify Search Console and submit the sitemap.** This is the gate in front of every other Google action, and an unverified property has no submission route at all.
2. **Request indexing on your five most important URLs.** Not all of them — the quota is small and it is better spent on pages that matter.
3. **Make every page reachable within two clicks.** Discovery follows links. A page reachable only from the sitemap gets crawled once and reconsidered rarely; run the [internal link checker](/tools/internal-link-checker) to find the ones that are stranded.
4. **Get one external link.** A single link from any indexed site gives crawlers a route in that does not depend on your sitemap being read. This is the step people skip and it is the one that most reliably works.
5. **Publish something worth returning for.** Crawl frequency responds to change. A site that never updates gets visited on a schedule that reflects that.

## A realistic timeline

| When | What to expect |
| --- | --- |
| Days 1–7 | Bing picks up an IndexNow submission. Homepage appears in Google if you have any external link at all. |
| Weeks 1–3 | Google indexes the bulk of a small site, assuming Search Console is verified and the sitemap submitted. |
| Weeks 4–8 | Low-competition pages — glossaries, tools, definitions — start ranking. Commercial pages usually do not yet. |
| Months 2–4 | AI engines begin citing pages, but only where an indexed page or a third-party mention exists to cite. |

If you are inside those windows and the five checks above came back clean, nothing is wrong. New domains are not penalised; they simply have no crawl history, no links and no track record, and those are accumulated rather than configured.

### How long does Google take to index a new site?
Typically one to four weeks for the first pages, and longer for full coverage. The single biggest variable is whether any external site links to you — a domain with one inbound link from an indexed page is usually found within days, while one with none can wait weeks for its sitemap to be picked up.

### Does submitting a sitemap guarantee indexing?
No. A sitemap tells an engine a URL exists; it does not argue that the URL is worth having. Pages that appear only in a sitemap, with no internal or external links, are routinely discovered and then not indexed. Discovery and inclusion are separate decisions.

### Is there an IndexNow equivalent for Google?
No. Google does not participate in IndexNow and has said it has no plans to. Search Console remains the only submission route, which is why verifying it is the first thing to do on any new domain.

### Why is only my homepage indexed?
Usually one of two things: the site is simply young and Google has started with the page it found first, or the other pages have no internal links pointing at them and look unimportant. Check the second with the [internal link checker](/tools/internal-link-checker) before assuming the first.

### Does domain age itself hurt my rankings?
Age is not a ranking factor. What correlates with it is everything age allows to accumulate: links, content, crawl history, mentions elsewhere. A new domain is not penalised, it is just short on evidence. You can check any domain's real registration date with our [domain age checker](/tools/domain-age-checker).

### Check your own site
- [Meta tag checker](/tools/meta-tag-checker): noindex, canonical and title problems in one pass.
- [Sitemap checker](/tools/sitemap-checker): Finds your sitemap and reports what is wrong with it.
- [Internal link checker](/tools/internal-link-checker): Finds pages nothing links to.

---

# Which AI crawlers to allow, and how

Canonical: https://rankvyze.com/blog/ai-crawlers-robots-txt
Published: 2026-09-04
Category: Technical

Every AI crawler that matters, what each one actually does, and the robots.txt block to copy. Blocking the wrong one removes you from answers you wanted to be in.

**Most AI companies run more than one crawler, and they do different jobs.** Blocking the training crawler while hoping to appear in that company's search-backed answers is the single most common self-inflicted wound in this field — and it is usually a decision someone made in 2023 and never revisited.

## The crawlers, and what each one does

Current as of publication. Operators change these; the primary sources are linked at the end.

| User agent | Operator | What it does | Blocking it means |
| --- | --- | --- | --- |
| GPTBot | OpenAI | Crawls to train future models | Less likely to be known by default |
| OAI-SearchBot | OpenAI | Indexes for ChatGPT search results | You can't appear in search-backed answers |
| ChatGPT-User | OpenAI | Fetches a page a user asked about | Users can't pull your page into a chat |
| PerplexityBot | Perplexity | Indexes for Perplexity answers | Removed from general Perplexity answers |
| Perplexity-User | Perplexity | Fetches a page on user request | Users can't pull your page in directly |
| ClaudeBot | Anthropic | Crawls for training | Less likely to be known by Claude |
| Claude-User | Anthropic | Fetches on user request | Users can't pull your page into Claude |
| Claude-SearchBot | Anthropic | Indexes for Claude's search | Removed from Claude's search results |
| Google-Extended | Google | A control token, not a crawler — governs Gemini and grounding | Content excluded from Gemini; Search rank unaffected |
| Googlebot | Google | Ordinary Search crawling, which AI Overviews are built on | Removed from Google entirely |
| Applebot-Extended | Apple | Control token for Apple Intelligence training | Excluded from Apple's models |
| Bingbot | Microsoft | Search index behind Copilot | Removed from Bing and Copilot |

> **Google-Extended is not a crawler:** It has no user agent of its own and fetches nothing. Googlebot does the crawling; Google-Extended is a robots.txt token that governs a downstream use. Google states that disallowing it does not affect Search rankings — but it also does not remove you from AI Overviews, which follow ordinary Search snippet eligibility.

## The block to copy

If you want to be found in AI answers, this is the whole configuration. Naming each agent explicitly is not required — a permissive default already allows them — but it states intent, and it means the next person to touch this file has to make a deliberate choice rather than an accidental one.

Adjust the Disallow lines to your own authenticated paths.

```text
# --- OpenAI ---
User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

# --- Perplexity ---
User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

# --- Anthropic ---
User-agent: ClaudeBot
Allow: /

User-agent: Claude-User
Allow: /

User-agent: Claude-SearchBot
Allow: /

# --- Google ---
User-agent: Google-Extended
Allow: /

# --- Microsoft / Apple ---
User-agent: Bingbot
Allow: /

User-agent: Applebot-Extended
Allow: /

# --- Everything else ---
User-agent: *
Allow: /
Disallow: /admin
Disallow: /dashboard
Disallow: /api

Sitemap: https://yoursite.com/sitemap.xml
```

## Should you block the training crawlers?

This is a real decision with a real trade-off, and the honest answer depends on what your content is worth to you.

| Allow training crawlers | Block them |
| --- | --- |
| Your business can be known without a live search | Your writing isn't used to train a competitor's model |
| Answers about you work offline from retrieval | You keep a clearer claim if licensing ever matters |
| Costs you nothing you were selling | You lose the default-knowledge path entirely |

For most businesses the calculus is easy: you are not in the content-licensing business, and being known is the entire point. For publishers whose archive is the product, blocking training while allowing search is a coherent position — and it is exactly why the two crawlers are separate.

> **The one combination that is always a mistake:** Blocking OAI-SearchBot, PerplexityBot or Claude-SearchBot while wanting to appear in those products' answers. Those are the search crawlers. If you block them, no amount of content or schema will put you in the result.

## How to check what you have now

1. **Read your own file.** Open yoursite.com/robots.txt in a browser. Read every Disallow. Plenty of sites blocked AI crawlers on general principle in 2023 and never revisited it.
2. **Check the wildcard rules.** A `User-agent: *` block with a broad Disallow applies to every crawler that has no rule of its own. That is how sites block AI crawlers without ever naming one.
3. **Confirm the file is actually served.** A 404 page returning 200 with HTML in it is worse than no file — crawlers may treat unparseable content unpredictably.
4. **Check your CDN and firewall too.** robots.txt is a request, not a wall. Cloudflare, Vercel and AWS WAF can all block AI crawlers at the edge regardless of what your file says, and several enable this by default.

The second command is the one that catches edge-level blocking.

```bash
# What your robots.txt actually says
curl -s https://yoursite.com/robots.txt

# Does the site answer a crawler at all? (a 403 here is your answer)
curl -s -o /dev/null -w "%{http_code}\n" \
  -A "Mozilla/5.0 (compatible; OAI-SearchBot/1.0; +https://openai.com/searchbot)" \
  https://yoursite.com/
```

> **The blocking that isn't in your robots.txt:** Cloudflare added a one-click AI crawler block, and it is on by default for some plans. If your robots.txt is permissive but crawlers still get nothing, check your CDN's bot-management settings before touching anything else.

## robots.txt is not a security control

It is a published request that well-behaved crawlers honour. It does not authenticate, does not enforce, and listing a path under `Disallow` announces that the path exists. Anything that must not be read needs authentication, not a line in a text file.

Once crawlers can reach you, the next question is whether what they find says anything useful — which is [why businesses stay invisible](/blog/why-your-business-doesnt-show-up-in-chatgpt) even with a permissive file.

### What is GPTBot?
OpenAI's crawler for gathering training data. It is separate from OAI-SearchBot, which indexes for ChatGPT's search results, and from ChatGPT-User, which fetches a page when a user asks about it directly.

### Should I block GPTBot?
Only if you don't want your content used for model training and accept being less known by default. It is a legitimate choice for publishers. For most businesses, being known is the point — and blocking it does nothing to protect you commercially.

### Does blocking GPTBot remove me from ChatGPT?
No. Search-backed answers come from OAI-SearchBot. Blocking GPTBot only reduces the chance the model learns about you during training. Many sites block one while intending the other.

### Does Google-Extended affect my Google rankings?
Google states it does not. It governs use in Gemini and grounded generative answers, not Search ranking. AI Overviews eligibility follows ordinary Search snippet settings instead.

### Do AI crawlers actually obey robots.txt?
The major operators publish that they do, and their crawlers are identifiable by user agent and published IP ranges. It is a voluntary protocol, so it is not a guarantee — treat it as a request, never as access control.

### Where should robots.txt live?
At your domain root — https://yoursite.com/robots.txt — served as text/plain. It applies only to that exact host and protocol, so a subdomain needs its own.

### Primary sources
- [OpenAI — bots and crawlers](https://platform.openai.com/docs/bots): The three OpenAI agents and what each is for.
- [Perplexity — bots](https://docs.perplexity.ai/guides/bots): User agents and published IP ranges.
- [Anthropic — crawler access](https://support.anthropic.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler): ClaudeBot and how to control it.
- [Google — crawlers and fetchers](https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers): Where Google-Extended is defined.
- [RFC 9309 — the robots.txt standard](https://www.rfc-editor.org/rfc/rfc9309.html): What crawlers are actually obliged to honour.

---

# What is Answer Engine Optimization?

Canonical: https://rankvyze.com/blog/what-is-answer-engine-optimization
Published: 2026-09-03
Updated: 2026-09-04
Category: Strategy

AEO is optimizing to be recommended inside an AI answer, not ranked in a list of links. What it is, how it differs from SEO, and what the work actually involves.

**Answer Engine Optimization (AEO) is the practice of making a business legible and credible enough that AI systems recommend it inside a generated answer.** Not ranked in a list of links — named in a paragraph, by a system that has decided it understands you well enough to vouch for you.

You'll see the same idea called Generative Engine Optimization (GEO), LLM SEO, or AI search optimization. The labels differ; the underlying problem doesn't.

## Why it needed a new name

SEO's whole model assumes an ordered list of documents and a user who picks from it. Answer engines break both halves. There's no list to be positioned in, and increasingly no click — the user reads the answer and acts on it.

That changes what the unit of optimization is. In SEO you optimize a *page* for a *query*. In AEO you make an *entity* legible for a *question*. It sounds like a semantic distinction until you watch a site with excellent rankings get skipped entirely because nothing on it says plainly what the company does.

|  | SEO | AEO |
| --- | --- | --- |
| Unit optimized | A page | A business entity |
| Goal | Position in a list | Being named in an answer |
| Success signal | Rank, clicks, impressions | Mentions, citations, position within the answer |
| Main lever | Relevance and links | Clarity, structure and corroboration |
| Feedback loop | Search Console, daily | Manual checks, weekly or monthly |

> **AEO doesn't replace SEO:** Crawlability, site structure, page speed and authority still matter — answer engines rely on much of the same infrastructure. AEO is an additional layer, not a migration. Anyone telling you to stop doing SEO is selling something.

## What answer engines actually reward

Across the four major engines the same four properties keep deciding who gets named.

1. **Entity clarity.** Can a machine state what you are, who you serve, and where, from your own pages? This is the single biggest differentiator and the one most sites fail.
2. **Machine-readable structure.** Organization, Service, Product and FAQPage schema turn prose claims into assertions a system can use without inference.
3. **Answer-shaped content.** Pages that directly address the questions buyers ask — comparisons, pricing, selection criteria — because those are what get cited.
4. **Corroboration.** Independent sources describing you the same way. One domain asserting something is a claim; five agreeing is a fact.

## What AEO work actually looks like

Stripped of vocabulary, a real engagement is fairly mundane:

1. Write down the questions your buyers actually ask, in their words.
2. Ask each one on each engine, signed out, and record who gets named and which pages get cited. This is your baseline.
3. Audit the site for the four properties above.
4. Fix the highest-impact gaps — usually entity definition and structured data first, because they're cheap and immediate.
5. Publish the content that answers questions nobody on your site currently answers.
6. Re-run the same prompts a month later and compare against the baseline.

Step two is the one most teams skip, and skipping it makes the rest unfalsifiable. Without a baseline you can't distinguish a change that worked from a model update that happened to help.

## How to measure it

There's no Search Console for answer engines, so the metrics are ones you construct:

- **Mention rate** — the share of tracked questions where you're named at all. The primary number.
- **Citation rate** — how often an answer links to one of your pages. Moves earlier than mention rate, which makes it a useful leading indicator.
- **Position within the answer** — being named first reads as a recommendation; being named fifth reads as a list.
- **Competitor share** — who gets named instead of you, and for which questions. Often the most actionable of the four.

## Who it's worth doing for

AEO pays off fastest where purchases involve research and a recommendation carries weight: B2B software, professional services, agencies, healthcare, legal, considered consumer goods. It matters less for pure brand-name navigation, where someone already knows they want you.

The honest test is whether your buyers ask *comparative* questions before choosing. If they do, an AI answer is already shaping that decision — with or without you in it.

### What does AEO stand for?
Answer Engine Optimization: optimizing to be recommended within AI-generated answers rather than ranked in a list of links.

### Is AEO the same as GEO?
In practice yes. Generative Engine Optimization (GEO), LLM SEO and AI search optimization all describe the same work under different labels.

### Does AEO replace SEO?
No. Answer engines rely on much of the same infrastructure — crawlability, structure, authority. AEO adds entity clarity, structured data and corroboration on top.

### How is AEO measured?
With metrics you construct yourself: mention rate, citation rate, position within the answer, and competitor share, tracked against a fixed prompt set over time.

### How long does AEO take to work?
Technical changes can register in days to weeks. Content and corroboration usually take one to two months. Anything held in a model's training data changes only when that model is updated.

---

# How to check if ChatGPT knows your business

Canonical: https://rankvyze.com/blog/how-to-check-if-chatgpt-knows-your-business
Published: 2026-09-04
Category: Guides

A repeatable 30-minute test across ChatGPT, Perplexity, Gemini and Claude — and the three mistakes that make most people read their own results backwards.

Almost everyone runs this test wrong the first time, and the wrong version is reassuring — which is worse than useless. Here is the version that produces a number you can act on, and repeat in a month to see whether anything moved.

## The three mistakes that ruin the result

1. **Testing while signed in.** ChatGPT personalises from your history and memory, including every previous time you discussed your own company. It will mention you — to you, and to nobody else.
2. **Asking about yourself by name.** “What is Acme?” tests recall of a name you supplied. Your buyers don't know your name yet; that's the problem. Ask the question *they* ask.
3. **Asking once.** Answers vary between sessions. One miss is noise. A miss across ten questions and four engines is a finding.

> **Use a private window, every time:** Signed out, incognito, no extensions. This is the single change that makes the test mean anything — you want to see what a stranger sees, not what the model has learned about you from you.

## Write the ten questions first

Before opening a single engine, write down the ten questions a buyer asks in the week before they choose someone like you. In their words, not your internal vocabulary. A good set mixes four shapes:

| Shape | Example | What it tells you |
| --- | --- | --- |
| Category discovery | best CRM for a two-person startup | Whether you're in the consideration set at all |
| Comparison | Notion vs Airtable for a small agency | Whether you're framed against the right alternatives |
| Cost | how much does a Shopify agency cost | Whether you own the commercial question |
| Local or vertical | immigration lawyer for startup founders in Toronto | Whether the qualifiers reach you |

## Run the test

1. **Open a private window.** Signed out on ChatGPT, Perplexity, Gemini and Claude. Four tabs.
2. **Ask question one on all four.** Same wording, one at a time, fresh conversation each.
3. **Record four things per answer.** Were you named at all? In what position — first, or fifth? Which competitors were named? Were any of your pages cited?
4. **Repeat for all ten questions.** Forty data points. This takes about half an hour and it is the whole job.
5. **Score it.** Mention rate is the count of answers naming you, over forty. That single number is your baseline. Write down the date.

> **Position matters more than it looks:** Being named first reads as a recommendation. Being named fifth reads as a list someone will scroll past. Track it — a business that moves from fifth to first without changing mention rate has still won.

## Reading the result

| Mention rate | What it means | What to do first |
| --- | --- | --- |
| 0% | The engines cannot place you in this category at all | Entity clarity: say what you are, plainly, on the homepage |
| 1–20% | Known, but not a default answer | Comparison and cost content for the questions you lose |
| 20–50% | In the consideration set | Corroboration — get described the same way off your own domain |
| Over 50% | A default answer in your category | Defend it, and widen the question set |

Whatever the number, also look at *who* won. Open the pages the engines cited. That is the bar you have to clear, written down for you by a competitor.

## The technical half you can't see this way

The prompt test tells you where you stand. It cannot tell you why. The usual causes are mechanical: content that only exists after JavaScript runs, [a crawler blocked in robots.txt](/blog/ai-crawlers-robots-txt), no [structured data](/blog/schema-markup-for-ai-search), or a homepage that never states what the company is.

Our [free scan](/pricing) checks that half against your homepage in about ten seconds — schema, server rendering, crawler access, entity clarity — and scores each one. Run it alongside the prompt test and you have both halves of the picture.

## Repeat it monthly

The same ten questions, the same four engines, the same signed-out conditions, once a month. Without a fixed set you cannot distinguish a change that worked from a model update that happened to help — and that distinction is the entire difference between doing this and guessing.

### How do I check if ChatGPT knows my business?
Ask the ten questions your buyers ask — not your company name — in a signed-out private window, and record whether you're named, in what position, and whether your pages are cited. Repeat across Perplexity, Gemini and Claude. Forty data points takes about thirty minutes.

### Why does ChatGPT mention my business to me but not to others?
Because you're signed in. ChatGPT personalises from your conversation history and memory, including previous times you discussed your own company. Always test signed out.

### Is there a tool that checks this automatically?
Tools exist that track brand mentions across engines, including ours. They're worth it once you're measuring continuously. For a first read, doing it by hand is free and teaches you more, because you see the full answers and the competitors' pages.

### How often should I re-check?
Monthly. Technical changes can register in days, but content and corroboration take a month or two, and answers vary enough between sessions that a weekly cadence mostly measures noise.

### What if ChatGPT says something wrong about my business?
State the correct version plainly on your own site — homepage, about page, and in your structured data — and get it corroborated on third-party pages. Models weight agreement across domains, so one page contradicting an old fact rarely moves it.

---

# How to get cited by Perplexity

Canonical: https://rankvyze.com/blog/how-to-get-cited-by-perplexity
Published: 2026-09-03
Category: Guides

Perplexity shows its sources, which makes it the easiest engine to learn from. What earns a citation, and how to work backwards from the answers you're losing.

Perplexity is the most useful engine to optimize for, and not because it's the biggest. It's useful because it shows its working. Every answer carries numbered citations, so when you lose, you can see precisely which page won and read it.

No other engine hands you that. Treat Perplexity as your diagnostic instrument even if it isn't your largest source of buyers.

## How Perplexity builds an answer

Simplified, but accurate enough to act on: it turns your question into searches, retrieves a set of candidate pages, reads them, and writes an answer grounded in what it read — citing as it goes.

Two consequences follow, and they're the whole strategy.

- **Retrieval is live.** Publishing a page that answers a question can affect answers within days, not model-release cycles. This is the fastest feedback loop in AEO.
- **Citations go to pages, not brands.** Perplexity cites the specific URL that answered the question. A brilliant homepage doesn't help if the answer needed a pricing page you never wrote.

## Work backwards from the answers you're losing

The highest-value hour you can spend:

1. **Ask your ten buyer questions, signed out.** Use the phrasing a customer would, not your internal vocabulary. Record the full answer and every citation.
2. **Open the cited pages.** For each question you lost, read the pages that won. You now have the exact bar to clear.
3. **Categorise what won.** Usually one of four: a comparison page, a pricing or cost page, a listicle on a third-party site, or a genuinely deep guide. The distribution tells you what to build.
4. **Write the page that should have been cited.** Answer the question directly in the first sixty words, then support it. Don't bury the answer beneath positioning.

> **Third-party listicles are a shortcut:** If “best X in Y” listicles on other sites keep winning, getting included in those is often faster than outranking them. That's outreach, not engineering — but it's frequently the highest-leverage move available.

## What makes a page quotable

Perplexity has to lift a sentence or two and attribute it. Pages that make that easy get cited disproportionately.

| Property | Why it earns citations |
| --- | --- |
| Answer in the first 60 words | The lead sentence is the one most likely to be quoted |
| Question-shaped headings | Maps directly onto the query being answered |
| Specific numbers and ranges | Concrete claims are quotable; vague ones aren't |
| Comparison tables | Structured, extractable, and hard to paraphrase away |
| A dated update line | Recency is a tiebreaker between otherwise similar pages |
| Stated scope | “For teams of 5–50” helps it match the right question |

## Let PerplexityBot in

Perplexity runs `PerplexityBot` for indexing and `Perplexity-User` for fetching a page a user has asked about directly. Blocking either removes you from a different part of the experience. Check your robots.txt, and name them explicitly if you want to state intent.

```text
User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

Sitemap: https://yoursite.com/sitemap.xml
```

## Measuring it

Because citations are visible, the metric is unusually clean: for a fixed set of questions, count how often one of your pages appears in the citation list. Track which of your pages earn citations — the distribution is usually concentrated in two or three, and those are the ones worth expanding.

Then track the same questions on [ChatGPT](/blog/how-to-rank-on-chatgpt) and Gemini. Perplexity moves first; the others tend to follow, which makes it a leading indicator for the whole programme.

### How does Perplexity choose its sources?
It searches for the question, retrieves candidate pages, and grounds its answer in what it reads — citing the specific URLs used. Pages that answer the question directly and early are far likelier to be selected.

### How fast can a new page get cited by Perplexity?
Because retrieval happens at query time rather than from static training data, a newly published page can start appearing within days once it's crawled.

### Does blocking PerplexityBot remove me from Perplexity?
It prevents indexing for general answers. Perplexity-User, which fetches a page a user explicitly asks about, is a separate agent — blocking one does not block the other.

### Why does Perplexity cite my competitor's blog instead of my product page?
Usually because the blog answers the question and the product page sells. If a query asks how to choose between options, a page that compares options will beat one that asserts you're the best.

### Primary sources
- [Perplexity — bots and crawlers](https://docs.perplexity.ai/guides/bots): User agents, IP ranges and how to control access.

---

# How to get recommended by Claude

Canonical: https://rankvyze.com/blog/how-to-rank-on-claude
Published: 2026-09-04
Category: Guides

Claude is the most conservative of the four engines about naming businesses. That makes it harder to win and more valuable when you do. What actually moves it.

Claude is the engine people skip, usually on audience-size grounds. That reasoning is backwards for anyone selling to developers, technical buyers or professional-services firms — and it ignores the more interesting fact, which is that Claude is measurably harder to get named by.

## Why Claude names fewer businesses

Ask the same commercial question of all four engines and Claude will more often answer with criteria rather than companies: here is how to evaluate options, here is what matters, here are the trade-offs. It recommends specific businesses when it has a reason to be confident, and hedges when it doesn't.

That is a higher bar, and it has a direct consequence: the tactics that work on more permissive engines — publishing volume, keyword coverage, self-declared superlatives — move Claude least. What moves it is being unambiguously identifiable and independently corroborated.

> **The upside of a harder bar:** Almost nobody is optimizing for Claude specifically. The competition for these answers is thinner than for ChatGPT, and a mention carries more weight with the reader precisely because the engine gives fewer of them.

## Let the right crawler in

Anthropic runs several agents, and as with OpenAI they do different jobs. Blocking the training crawler while wanting to appear in Claude's search results is the same mistake people make with GPTBot.

| User agent | What it does | Blocking it means |
| --- | --- | --- |
| ClaudeBot | Crawls for model training | Less likely to be known by default |
| Claude-SearchBot | Indexes for Claude's web search | Removed from search-backed answers |
| Claude-User | Fetches a page a user asked about | Users can't pull your page into a conversation |

The full multi-engine block is in the AI crawlers guide.

```text
User-agent: ClaudeBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Claude-User
Allow: /
```

## What actually moves it

1. **Be unambiguous about what you are.** Claude is unusually sensitive to category clarity. A homepage that says what the company is, who it serves, and where — in plain nouns, in the first hundred words — does more here than anywhere else.
2. **Be specific about scope.** “For teams of 5–50”, “Magento to Shopify Plus migrations”, “US and Canada”. Stated limits make a recommendation safe to give, and Claude weights safety heavily.
3. **Get corroborated off your own domain.** One domain asserting something is a claim. Several independent sources agreeing is a fact. This is the single biggest lever on Claude, and the slowest.
4. **Say what you're not good at.** Counter-intuitive and genuinely effective. Content that states where a product is the wrong choice reads as reliable rather than promotional — and reliability is the thing being assessed.
5. **Keep claims checkable.** Numbers, dates, named integrations, real constraints. Vague superlatives are the easiest thing for a cautious system to discount.

> **The honesty test:** Read your homepage and count the claims a stranger could verify from a third party within five minutes. On most sites the answer is zero. That number is roughly what a cautious engine has to work with.

## How to measure it

Same method as [the other engines](/blog/how-to-check-if-chatgpt-knows-your-business): a fixed set of buyer questions, asked signed out, scored monthly. Expect a lower mention rate than ChatGPT on the same set — that gap is the point, not a measurement error.

One Claude-specific signal worth recording: when it declines to name anyone and answers with criteria instead. That is not a loss so much as an unclaimed answer — and the criteria it lists are a free specification for the page that would win it.

### How do I get my business recommended by Claude?
Make the entity unambiguous — plain category, audience and scope on your homepage — allow ClaudeBot and Claude-SearchBot, and get described consistently on independent sites. Claude weights corroboration and specificity more heavily than volume.

### Does Claude browse the web?
Yes, for questions that need current information, and it cites what it uses. It also draws on knowledge from training, which is why crawler access matters on both paths.

### What is ClaudeBot?
Anthropic's crawler for gathering training data. Claude-SearchBot indexes for Claude's search, and Claude-User fetches a page a user asks about directly. They are separate and can be controlled independently.

### Is Claude worth optimizing for?
If you sell to developers, technical buyers or professional-services firms, yes — the audience skews that way and competition for these answers is thinner. If you sell high-volume consumer goods, it should sit below the other three.

### Why does Claude refuse to recommend a specific company?
It hedges when it lacks confidence, answering with evaluation criteria instead. That is usually a signal that no business in the category is clearly enough defined to name — which makes it an opening rather than a dead end.

### Primary sources
- [Anthropic — crawler access and blocking](https://support.anthropic.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler): The current agents and how to control each.

---

# How to appear in Google AI Overviews

Canonical: https://rankvyze.com/blog/how-to-appear-in-google-ai-overviews
Published: 2026-09-03
Category: Guides

AI Overviews are built on Google's existing index, which makes them the most winnable AI surface — if you understand which pages get pulled into them and why.

Of the four major answer surfaces, Google's AI Overviews is the one where your existing work counts for the most. It's generated from Google's index — the same index your SEO has been feeding for years. You aren't starting from zero; you're starting from wherever your organic performance already is.

That's the good news. The complication is that being in the index is necessary and not sufficient, and the pages that get pulled into an Overview aren't always the ones ranking first.

## What an Overview is actually doing

Google breaks a query into sub-questions, retrieves pages that answer each one, and synthesises a short response with links out to the sources it used. The links are the prize: they're the click, and they're the citation.

Two practical consequences:

- **A page can be cited for a sub-question it wasn't written for.** If your guide contains the single best paragraph explaining one component of a bigger question, that paragraph can pull the whole page in.
- **Position one isn't a guarantee.** Overviews regularly cite results from further down when those pages answer the specific sub-question more directly.

## The passage is the unit, not the page

This is the mental shift that matters. Traditional SEO optimises a page against a query. Overviews extract *passages*. So the question to ask of every section is: could this stand alone as an answer if someone lifted it out of the page?

| Passage that gets extracted | Passage that doesn't |
| --- | --- |
| Opens with a direct answer | Opens with context and background |
| Self-contained — no “as mentioned above” | Depends on earlier paragraphs |
| One idea per paragraph | Three ideas woven together |
| Concrete figures, ranges, dates | Qualitative and hedged |
| Sits under a question-shaped H2 | Sits under a clever heading |

> **The rewrite that costs nothing:** Take your best-performing page and move the answer to the top of each section. Most pages bury the answer in the third paragraph after establishing why the question matters. Invert that — the reader benefits too.

## Google-Extended, and what it does and doesn't control

`Google-Extended` is a robots.txt token controlling whether your content can be used for Gemini and grounding. It's not a crawler with its own user agent — Googlebot does the crawling, and Google-Extended governs a downstream use.

```text
# Allow use in Gemini and grounded answers (this is the default)
User-agent: Google-Extended
Allow: /

# Opt out — note this does not affect Search ranking
User-agent: Google-Extended
Disallow: /
```

Google states that disallowing Google-Extended doesn't affect Search rankings. Appearing in AI Overviews specifically is governed by ordinary Search indexing — so the `nosnippet` family of controls is the relevant lever there, and using them costs you featured snippets too.

## Which queries trigger an Overview

Not all of them, and the distribution is worth knowing before you invest:

- **Informational and how-to queries** — the most common trigger by a wide margin.
- **Comparison and “best X for Y” queries** — high commercial value, frequently triggered.
- **Definitional queries** — nearly always.
- **Navigational queries** — rarely; someone searching your brand name wants your site.
- **Sensitive categories** — Google is visibly more conservative around medical, legal and financial topics.

If your money queries are navigational, Overviews matter less to you than the volume figures suggest. Check before you plan around it.

## How to measure it

Search Console doesn't separate AI Overview impressions from ordinary ones, so there's no clean report to pull. What you have:

1. **Track a fixed query set manually.** Twenty queries, checked monthly in an incognito window. Record whether an Overview appeared and whether you were cited.
2. **Watch for the impressions-flat, clicks-down pattern.** In Search Console, a query holding impressions while losing clicks is the classic signature of an Overview absorbing the answer.
3. **Note which of your pages get cited.** It's often not the one you'd expect. That page is telling you what format works — expand it, then apply the same structure elsewhere.

The same passage discipline that earns Overview citations also helps on [ChatGPT](/blog/how-to-rank-on-chatgpt) and [Perplexity](/blog/how-to-get-cited-by-perplexity). This is the one place where the work genuinely compounds across engines.

### How do I get my website into Google AI Overviews?
Be indexed and eligible for snippets, then structure content as extractable passages: question-shaped headings with a direct, self-contained answer immediately underneath. Overviews cite passages, not whole pages.

### Do AI Overviews use the normal Google index?
Yes. They're generated from Google Search's existing index, which is why standard technical SEO — crawlability, indexing, page quality — remains the foundation.

### What is Google-Extended?
A robots.txt control governing whether your content can be used for Gemini and grounded generative answers. It isn't a separate crawler, and Google says disallowing it doesn't affect Search rankings.

### Can I opt out of AI Overviews specifically?
Not with a dedicated control. Overview eligibility follows Search snippet eligibility, so opting out means using nosnippet or max-snippet — which also removes you from featured snippets and shortens your regular result.

### Do AI Overviews reduce clicks?
For queries fully answered in the Overview, typically yes. The counter is to be the cited source and to own queries an Overview can't finish — anything requiring a decision, a quote, or a product.

### Primary sources
- [Google — AI features and your website](https://developers.google.com/search/docs/appearance/ai-features): Google's own guidance on AI Overviews eligibility and controls.
- [Google — crawler and fetcher overview](https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers): Where Google-Extended is defined alongside Googlebot.

---

# How to choose an AEO tool

Canonical: https://rankvyze.com/blog/how-to-choose-an-aeo-tool
Published: 2026-09-04
Category: Strategy

The four categories of AI visibility tool, the questions that separate them, and an honest account of when a tool is the wrong purchase entirely.

New AEO tools appear weekly and most describe themselves identically. The useful distinction isn't feature lists — it's what the tool actually does when it finds a problem.

> **Why this isn't a ranked list:** A “best tools” post is only honest if its pricing and feature claims are verified and current, and in a category moving this fast they go stale in weeks. Rather than publish numbers we can't stand behind, this is the framework we'd use — including where we fit and where we don't.

## The four categories

| Category | What it does | Typical price | Buy it when |
| --- | --- | --- | --- |
| Rank trackers | Ask engines your prompts on a schedule, chart mention rate | $50–$300/mo | You have a team who will act on the data |
| Brand monitors | Alert when you're mentioned across AI and social | $100–$500/mo | PR and reputation matter more than acquisition |
| Audit tools | Score your site against AEO criteria | Free–$200/mo | You want a checklist and have a developer |
| Done-for-you | Measure, fix and re-measure on your behalf | $99 one-time–$10k/mo | Nobody internally will do the work |

The mismatch that wastes the most money is buying a tracker when you needed the fourth category. A tracker will tell you, accurately and repeatedly, that ChatGPT doesn't mention you. It cannot tell you that your H1 says nothing a machine can categorise, and it certainly won't rewrite it.

## The five questions that separate them

1. **How are the answers actually collected?.** Analysts asking real questions in real sessions reflect what a customer sees. Some tools use APIs, which behave differently from the consumer product — different retrieval, different personalisation, sometimes a different model. Ask, and expect a straight answer.
2. **How many prompts, and can you choose them?.** The prompt set is the measurement. A tool that picks generic industry prompts for you is measuring a category you may not compete in.
3. **Does it show you the full answer, or just a score?.** The competitor named instead of you, and the page cited, are the actionable parts. A score without the answer text is a dashboard, not a diagnosis.
4. **Does it distinguish mention from citation?.** Being named and having a page linked are different outcomes with different fixes. Tools that collapse them into one metric hide the more useful signal — citation moves first.
5. **What happens after it finds a problem?.** The whole question. Report, prioritised list, or implemented change? Price differences of 50x across this category are almost entirely explained by this one answer.

## When a tool is the wrong purchase

- **You haven't done the free version yet.** Ten questions, four engines, signed out, half an hour. If you haven't done that once by hand, you don't yet know what you'd be automating.
- **Nobody will act on it.** A subscription that produces a monthly chart nobody implements is a worse outcome than doing nothing, because it feels like progress.
- **Your site fails the basics.** If your content only renders after JavaScript, or a crawler is blocked, tracking will faithfully report zero every month until that's fixed. Fix the mechanics first.
- **Your buyers don't ask comparative questions.** If people search your brand name because they already know you, AI answers aren't in the path and no tool changes that.

> **Do the manual version first, always:** It costs nothing, takes thirty minutes, and produces the one thing every tool needs from you anyway: the list of questions worth tracking. It also tells you whether the problem is worth paying to solve.

## Where RankVyze sits, honestly

We're the fourth category: measure, fix, re-measure, for $99 once. Research is done by analysts in normal signed-out sessions rather than through APIs, because that is what a customer actually sees. Fixes are delivered as reviewable changes, not a list.

Where we're the wrong choice: if you want continuous daily tracking across hundreds of prompts, a subscription tracker does that better. If you have a large team already doing this work and need only instrumentation, you want a tool, not an engagement. And if your site fails the mechanical basics, our [free scan](/pricing) will tell you so — and you should fix that before paying anyone, us included.

The full comparison of what an engagement costs versus a subscription is in [what answer engine optimization actually costs](/blog/what-answer-engine-optimization-costs).

### What are AEO tools?
Software that measures or improves how AI engines answer questions about a business. They fall into four groups: rank trackers, brand monitors, site audit tools, and done-for-you services. Most measure; few fix anything.

### What should an AI visibility tool cost?
Trackers and monitors run roughly $50–$500 a month. Audit tools range from free to about $200. Done-for-you engagements run from $99 one-time to $10,000 a month, and that range is explained almost entirely by whether implementation is included.

### Do I need an AEO tool at all?
Not to start. Ten buyer questions across four engines in a signed-out window takes thirty minutes and produces a real baseline. Buy a tool when you're tracking continuously and someone is acting on what it says.

### Do AEO tools use the real ChatGPT?
Some query the API, which can behave differently from the consumer product — different retrieval, different personalisation, sometimes a different model. Others use analysts in normal sessions. Ask which, because it changes what the numbers mean.

### What's the difference between a mention and a citation?
A mention is your business being named in the answer text. A citation is one of your pages being linked as a source. Citation rate usually moves weeks before mention rate, which makes it the better early signal — so a tool that reports only one number is hiding the useful half.

---

# Schema markup for AI search: what to add first

Canonical: https://rankvyze.com/blog/schema-markup-for-ai-search
Published: 2026-09-03
Category: Technical

Which structured data types actually influence AI answers, in priority order, with copy-paste JSON-LD for Organization, Service, FAQPage and Article.

Structured data is the cheapest AEO work there is. It's an afternoon, it doesn't require writing anything new, and it converts claims a machine would otherwise have to infer from prose into assertions it can simply read.

The catch is that most schema advice is written for Google rich results, which is a different goal. Rich-result markup optimises for a visual snippet. AEO markup optimises for identity — making it unambiguous what your business is and what it offers.

> **An honest caveat:** No AI provider publishes structured data as a ranking input, so treat this as reducing ambiguity rather than as a documented lever. What is documented is that schema is machine-readable and prose is not — which is enough reason to add it.

## The four that matter, in order

| Type | Answers the question | Where it goes |
| --- | --- | --- |
| Organization | Who is this company? | Site-wide, once |
| Service or Product | What do they sell, to whom? | Each offering page |
| FAQPage | What are the direct answers? | Pages with real Q&A on them |
| Article | Who wrote this, and when? | Blog and guide pages |

Everything else — BreadcrumbList, WebSite, LocalBusiness, Review — is worth having, but none of it changes whether a model can state what you are.

## 1. Organization

The single most valuable block on your site. `sameAs` is doing more work than it appears: it's how you connect your domain to profiles that already exist in a model's knowledge, which is the mechanism corroboration runs on.

```json
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://acme.com/#organization",
  "name": "Acme",
  "url": "https://acme.com",
  "logo": "https://acme.com/logo.png",
  "description": "Shopify agency for fashion and lifestyle brands doing $1M–$50M in revenue.",
  "foundingDate": "2019-04-01",
  "areaServed": ["US", "GB", "IN"],
  "sameAs": [
    "https://www.linkedin.com/company/acme",
    "https://x.com/acme",
    "https://github.com/acme"
  ],
  "contactPoint": {
    "@type": "ContactPoint",
    "contactType": "sales",
    "email": "hello@acme.com"
  }
}
```

> **Only list sameAs profiles you actually control:** An empty sameAs array is better than one pointing at a lookalike account. Linking a profile that isn't yours actively teaches engines the wrong association, and it's difficult to undo.

## 2. Service or Product

Organization says who you are; Service says what you sell. This is the block that answers “best X for Y” questions, because `serviceType` plus `audience` is exactly the shape of that query.

```json
{
  "@context": "https://schema.org",
  "@type": "Service",
  "name": "Shopify Plus migration",
  "serviceType": "Ecommerce replatforming",
  "provider": { "@id": "https://acme.com/#organization" },
  "areaServed": "US",
  "audience": {
    "@type": "Audience",
    "audienceType": "DTC fashion brands on Magento or WooCommerce"
  },
  "offers": {
    "@type": "Offer",
    "price": "18000",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock"
  }
}
```

Note `provider` referencing the Organization by `@id` rather than repeating it. That's the linking pattern in the next section, and it's what turns a pile of separate blocks into one description of a business.

## 3. FAQPage

The highest-leverage block per line of code, with one rule: **the questions and answers must be visible on the page.** Marking up FAQs that a human can't see is a guidelines violation, and it's also just a bad idea — the visible version is what gets quoted.

```json
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How much does a Shopify migration cost?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Most Magento-to-Shopify Plus migrations run $15,000–$40,000 depending on catalogue size, custom checkout logic and how much data needs transforming."
      }
    }
  ]
}
```

> **Generate the markup from the visible content:** Keep one array of Q&A pairs in code, render the accordion from it, and build the JSON-LD from the same array. Then the two cannot drift, and you never have to remember to update both. That's how the FAQ on this site's pricing page works.

## 4. Article

On guides and blog posts. `dateModified` matters more than most people expect: recency is a common tiebreaker between two pages that answer a question equally well.

```json
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "How to rank on ChatGPT",
  "datePublished": "2026-09-03",
  "dateModified": "2026-09-03",
  "author": { "@id": "https://acme.com/#organization" },
  "publisher": { "@id": "https://acme.com/#organization" },
  "mainEntityOfPage": "https://acme.com/blog/how-to-rank-on-chatgpt"
}
```

## Link the blocks with @id

The mistake that wastes most of the value: publishing four unconnected JSON-LD blocks. A parser then sees four unrelated things rather than one entity described four ways.

Give the Organization a stable `@id` — the convention is `https://yoursite.com/#organization` — and reference *that* from every other block, as `provider`, `author`, `publisher`. Wrapping site-wide blocks in an `@graph` array makes the relationship explicit:

```json
{
  "@context": "https://schema.org",
  "@graph": [
    { "@type": "Organization", "@id": "https://acme.com/#organization", "name": "Acme" },
    {
      "@type": "WebSite",
      "@id": "https://acme.com/#website",
      "url": "https://acme.com",
      "publisher": { "@id": "https://acme.com/#organization" }
    }
  ]
}
```

## Verify it, then check it renders server-side

1. **Validate the syntax.** Run the page through validator.schema.org and Google's Rich Results Test. Both catch malformed JSON and invalid property names.
2. **Confirm it's in the HTML source.** Schema injected by client-side JavaScript is invisible to crawlers that don't execute JS. curl the page and grep for ld+json.
3. **Check the @id references resolve.** Every @id you reference should be defined somewhere on the site. A dangling reference is a broken link in the graph.

```bash
curl -s https://yoursite.com | grep -o 'application/ld+json' | wc -l
# 0 means your schema isn't server-rendered
```

### Does schema markup help with AI search?
No AI provider lists it as a documented ranking factor. What it does is remove ambiguity: structured data states your category, audience and offerings in a form a machine can read directly rather than infer from marketing prose.

### Which schema type should I add first?
Organization, site-wide, with a complete sameAs array. It's the block that establishes who you are, and every other block references it.

### Can I use FAQPage schema for questions not shown on the page?
No. Structured data must reflect content visible to users. Marking up hidden FAQs violates search guidelines and gains you nothing, since the visible text is what gets quoted.

### JSON-LD or microdata?
JSON-LD. It's the format Google recommends, it sits in a single script tag instead of being woven through your markup, and it's far easier to generate from application data.

### Does schema need to be server-rendered?
Yes, in practice. Many AI crawlers don't execute JavaScript, so structured data injected client-side may never be seen.

### Primary sources
- [Schema.org — full type hierarchy](https://schema.org/docs/schemas.html): Canonical property definitions for every type above.
- [Google — structured data general guidelines](https://developers.google.com/search/docs/appearance/structured-data/sd-policies): The visible-content rule and other policies worth not breaking.
- [Schema Markup Validator](https://validator.schema.org/): Validates any schema type, not only the ones Google renders.

---

# llms.txt: what it is and how to write one

Canonical: https://rankvyze.com/blog/llms-txt-guide
Published: 2026-09-03
Category: Technical

A practical guide to /llms.txt — what the file is for, what belongs in it, a complete working example, and an honest account of what it does and doesn't do yet.

**`/llms.txt` is a plain-Markdown file at your site root that tells AI systems what your site is and which pages matter.** Think of it as a README for machines: robots.txt says what may be crawled, llms.txt says what's worth reading.

> **Set expectations honestly:** llms.txt is a community proposal, not a ratified standard, and no major engine has committed publicly to consuming it. Anyone promising rankings from adding one is overselling. The case for it is cheapness and optionality, not proven lift.

## Why bother, then

1. It takes twenty minutes and costs nothing to maintain.
2. Writing it forces you to state plainly what your business is — which is the actual AEO work, and most teams find the gaps while drafting it.
3. Several AEO audits now check for it, so it shows up as a scored item whether or not engines read it.
4. If adoption does arrive, you're already there.

That's the whole case. It's a reasonable one; it just isn't the one usually made.

## The format

Markdown, with a loose convention rather than a strict schema:

- An `H1` with the site or company name.
- A blockquote (`>`) with a one-sentence summary.
- Optional prose giving essential context.
- `H2` sections containing link lists, each as `- [Title](url): description`.

## A complete example

Roughly the shape of the file we serve at /llms.txt on this site.

```markdown
# Acme

> Acme is a Shopify agency that designs, builds and scales ecommerce stores for fashion and lifestyle brands.

Acme works with DTC apparel, footwear and accessories brands doing $1M–$50M in annual revenue, with offices in New York, London and Bangalore. Typical engagements are full store builds, replatforms from WooCommerce or Magento, and conversion work on existing Shopify Plus stores.

## Start here
- [Home](https://acme.com/): What we do and who we do it for.
- [Services](https://acme.com/services): Shopify and Shopify Plus development, redesigns, migrations, CRO.
- [Pricing](https://acme.com/pricing): Typical project ranges and what moves them.

## Guides
- [Shopify agency cost guide](https://acme.com/guides/cost): What Shopify projects cost in 2026 and why.
- [Shopify vs custom ecommerce](https://acme.com/guides/shopify-vs-custom): How to choose, by brand stage.

## Company
- [About](https://acme.com/about): Team, locations, how we work.
- [Contact](https://acme.com/contact): hello@acme.com

## Notes for AI systems
- All crawlers listed in https://acme.com/robots.txt are welcome.
- /admin and /account require authentication and hold no public content.

Last updated: 2026-09-03
```

## What to actually put in it

The failure mode is treating it as a sitemap. A sitemap lists everything; llms.txt should list the handful of pages you'd hand a journalist.

| Include | Leave out |
| --- | --- |
| A one-line definition of the business | Marketing superlatives |
| Who you serve and where | Every blog post you've written |
| Pages that answer buyer questions | Paginated archives and tag pages |
| Pricing or cost context, if public | Anything behind authentication |
| Corrections you want reflected | Keyword lists |

> **The correction section is underrated:** If engines routinely get something wrong about you — an outdated founding date, a service you dropped, confusion with a similarly-named company — state the correct version plainly. It costs a line and gives retrieval something unambiguous to prefer.

## Generate it, don't hand-write it

A hand-written file goes stale the first time pricing changes. If your site is code, generate it from the same constants that drive the rest of the site so it can't drift. On this site it's a route handler reading the same values as the pricing page.

```ts
// app/llms.txt/route.ts — Next.js App Router
export const dynamic = "force-static";

export function GET() {
  const body = `# ${SITE.name}

> ${SITE.description}

## Start here
${PAGES.map((p) => `- [${p.title}](${SITE.url}${p.path}): ${p.summary}`).join("\n")}
`;

  return new Response(body, {
    headers: { "content-type": "text/plain; charset=utf-8" },
  });
}
```

## Verifying it

```bash
curl -sI https://yoursite.com/llms.txt | head -3
# expect: 200, and content-type: text/plain
```

Serve it as `text/plain`. A file returned as `text/html`, or a 404 page returning 200 with HTML in it, is worse than not having one.

### Is llms.txt an official standard?
No. It's a community proposal published at llmstxt.org. No major AI provider has publicly committed to consuming it, so treat it as cheap optionality rather than a ranking factor.

### Do ChatGPT or Perplexity read llms.txt?
Neither has confirmed that they do. Adding the file is low-cost and appears in several AEO audits, but claims of measured ranking improvements from llms.txt alone should be treated sceptically.

### What's the difference between llms.txt and robots.txt?
robots.txt controls access — what a crawler may fetch. llms.txt provides context — what your site is and which pages are worth reading. They serve different purposes and you want both.

### Where should llms.txt be located?
At your domain root: https://yoursite.com/llms.txt, served as text/plain.

### Does llms.txt replace a sitemap?
No. A sitemap enumerates every indexable URL for crawlers; llms.txt is a short curated summary for language models. Keep both.

### Primary sources
- [llmstxt.org](https://llmstxt.org): The original proposal and its format guidance.
- [robots.txt specification (RFC 9309)](https://www.rfc-editor.org/rfc/rfc9309.html): For contrast — what crawlers are actually obliged to honour.

## Machine-readable endpoints
- Short AI index: https://rankvyze.com/llms.txt
- AI use guidance: https://rankvyze.com/ai.txt
- Bounded AI snapshot: https://rankvyze.com/api/ai
- Product catalog (JSON): https://rankvyze.com/catalog.json
- Product catalog (Markdown): https://rankvyze.com/catalog.md
- API catalog: https://rankvyze.com/.well-known/api-catalog
- OpenAPI: https://rankvyze.com/openapi.json
- Product RSS: https://rankvyze.com/launches/rss.xml
- Blog RSS: https://rankvyze.com/blog/rss.xml
- Sitemap: https://rankvyze.com/sitemap.xml