Anthropic

How to get your business mentioned in Claude.

Choose the Claude experience you intend to measure before calling something a rank. A conversation with web search, a conversation without it and an API call with specified tools are different checks. Useful source material and recorded context matter more than a single impressive screenshot.

By the RankVyze team · Updated September 19, 2026

How do I improve my visibility in Claude?

Review access for Claude-SearchBot and user-requested fetching, publish clear and verifiable information about your business, and answer the relevant buyer question thoroughly. Measure with a saved prompt and an explicit search setting. Keep source citations separate from brand recommendations. An API check can document one response, but it cannot establish your position across the Claude app or other engines.

The work

A practical workflow for Claude.

In order. Each step is a thing you can check off, not a principle to keep in mind.

1
Review each crawler’s purpose
Anthropic documents ClaudeBot for training, Claude-SearchBot for search and Claude-User for user requests. Decide access by purpose and investigate relevant firewall rules as well as robots.txt.
2
Document what your product actually does
Use concrete descriptions, supported use cases and current limitations. Distinguish features available now from a roadmap item or a manual workflow.
3
Make evidence accessible
Publish important facts on public pages with clear headings and references. A source that explains its methodology lets a reader assess a claim instead of relying on a badge or score alone.
4
Check an unprompted discovery question
Use a question a real buyer would ask without naming you. Keep branded questions in a separate group so the resulting mention rates mean something.
5
Retain the full check context
Save model or product, tools, prompt, timestamp, response and sources. Label incomplete responses as incomplete instead of interpreting them as evidence that your business is absent.

Mechanism

How Claude builds an answer.

01
Anthropic provides separate controls for training, search discovery and user-initiated retrieval. A training choice should not be mistaken for a complete search-access policy.
02
Inspect the source links when a response uses web search. They show evidence from that response, not the model’s full knowledge of your company.
03
In your own reporting, distinguish facts in the cited page from inferences in the answer. Correct inaccurate public product information before trying to amplify it.

robots.txt

Which crawlers to allow.

Search discovery, user-requested fetching and model training can use different controls. Choose access by purpose and check the official documentation below.

Claude-SearchBot

Search discovery to improve relevance and accuracy of search results.

Claude-User

Page retrieval requested by a user.

ClaudeBot

Crawling for model training, distinct from search and user requests.

Our recommendations

Where to focus your Claude work.

Explain a reproducible workflow

A walkthrough with real inputs, outputs and limitations is stronger evidence of what a product does than a generic benefit statement. Keep demonstrations labelled if they use sample data.

Keep current features separate from plans

For RankVyze, analyst-entered customer research and a future integration are different capabilities. Clear availability information helps prevent a source from describing a product that does not yet exist.

Clarify similar product names

Give readers a consistent domain and category alongside the brand name. Make the About and contact pages easy to find, and avoid claiming profiles that have not been created.

Preserve the original answer

A yes/no mention flag loses context. Keep the text so a reviewer can see whether the name was recommended, rejected, compared or merely repeated from the question.

Common mistakes

Claiming Claude ranks sources in a known fixed order.
Reporting an API result as if it were a consumer-app search position.
Calling a disabled integration live or displaying invented results.
Promising a website edit will immediately change model training knowledge.

How fast it moves

No universal timeline establishes when Claude will discover, cite or recommend a particular page. Date your changes and compare repeated checks with matching settings. Report uncertainty instead of converting a service deadline into a model behaviour claim.

FAQ

Claude questions.

Does allowing ClaudeBot guarantee a mention?

No. ClaudeBot is documented as a training crawler. Search and user requests have separate agents, and permission to access a page does not guarantee that an answer selects it.

What does the RankVyze AI rank checker measure?

It is designed for a Claude API web-search answer snapshot, with its current availability shown on the tool page. It does not measure ChatGPT, Google positions or all Claude app responses.

Should I put my brand in the test prompt?

For a discovery baseline, use a natural buyer question without your name. You can also test branded questions, but label them separately because the prompt itself supplies the brand.

Sources and scope

These official references support the technical controls and reporting described here. The workflow is RankVyze’s editorial advice; it does not reveal proprietary ranking weights or guarantee inclusion.