How to check your Claude visibility (free tool)

Checking your Claude visibility means testing two things: whether Claude recalls your brand from training, and whether its web search can reach and cite your live pages. You can test the first with a few careful prompts and the second with a structured audit. Here is the full process, step by step.

Key takeaway: Test Claude’s model-memory and live-web behaviour separately. For live visibility, check Claude-SearchBot and Claude-User access, then compare the sources Claude actually uses. ClaudeBot is a separate training choice, not a live-search requirement. The free Claude Visibility Checker audits the website readiness layer.

Why check Claude visibility separately from ChatGPT?

Each AI engine sources its answers differently, so a check built for one quietly lies to you about the others. Claude answers from its own training data by default, and when its web search capability is enabled (in claude.ai, the Anthropic API and Claude Code) it fetches pages directly rather than reading Google’s index. Anthropic also runs its own named crawlers. That means you can be visible in ChatGPT and absent from Claude, or the reverse, and a Google rankings report will not show you either way.

Claude also reaches people in places a manual test cannot see. A large share of Claude usage happens through the API, inside other companies’ assistants and internal tools, and through Claude Code on developers’ machines. You cannot test those surfaces one by one, which is why the technical signals underneath them are what a structured check should measure.

Step 1: test what Claude recalls from training

Open claude.ai with web search switched off (or simply note when Claude answers without citing sources) and ask questions a real buyer would ask:

• “What can you tell me about [your company name]?”
• “What are the best [your product or service category] options in [your location or niche]?”
• “Recommend a [type of business] that specialises in [your specialism].”

Record three things for each answer: whether your brand is named at all, whether the description is accurate, and which competitors appear instead of you. If Claude describes you accurately without searching, your training-era footprint is working. If it has never heard of you, or confuses you with a similarly named company, your entity signals need work: that is a slow-burn fix built on consistent third-party references.

Step 2: test what Claude cites live

Now repeat the same questions with web search enabled. This tests a completely different system: Claude searching the web, opening pages and quoting what it can read right now.

Watch for the difference between being found and being cited. Claude may open your site but quote a competitor because their page states the answer in one liftable sentence while yours buries it. If your brand appears with a source link to your own domain, note which page Claude chose: that tells you what it considers your most quotable content.

Step 3: check whether Anthropic’s crawlers can reach you

Manual prompting tells you whether you appear, but not why not. The first technical check is robots.txt. Visit yourdomain.com/robots.txt and look for rules affecting these three agents:

• Claude-SearchBot, used for search
• Claude-User, which fetches a page when somebody asks Claude about it
• ClaudeBot, which collects content for model training

Anthropic says all three honour robots.txt, but they do different jobs. A block on Claude-SearchBot or Claude-User can affect live retrieval; a ClaudeBot block is a training-policy decision. SearchScore’s July 2026 interactive sample found 6.9% of sites blocked at least one major AI-related crawler or control checked by the audit.

Step 4: check whether your content is readable and quotable

Even with crawler access, two structural problems stop citations:

JavaScript-only content. If your key content only appears after client-side JavaScript runs, a crawler fetching your HTML sees an empty page. View your page source (not the rendered inspector) and confirm your main copy is actually in it.

Unclear or incomplete answers. Check whether important facts are easy to find and supported. Self-contained passages can improve readability and retrieval, but Anthropic does not publish a rule that Claude selects whichever source has the cleanest quotable sentence. The historical SearchScore structure benchmark is a readiness measure, not a citation explanation.

Step 5: run a structured audit

Manual checks measure the actual answer. The free Claude Visibility Checker audits the underlying website readiness: Anthropic agent access, renderability, content clarity and entity consistency. It cannot inspect Claude’s training data or predict which source Claude will choose, so pair it with repeated buyer prompts.

A useful habit: run it on your own domain, then on the competitor Claude keeps recommending instead of you. The gap between the two scores is usually the clearest brief you will ever get for what to fix first.

How to read your results

A low score is rarely about content quality. The report separates two failure modes so you do not waste effort:

• “Claude can’t reach or read you.” An access or rendering issue: a crawler is blocked, or your content only exists after JavaScript runs. High impact, often a fast fix, and it caps everything else.
• “Claude can reach you but has no reason to pick you.” A signal issue: weak entity data, no citable passages, or a thin third-party footprint. Slower to move, but it decides whether you are named versus merely readable.

Fix a genuine Claude-SearchBot or Claude-User access failure early when it exists. If live access is already healthy, move straight to the cited-source and content comparison rather than treating crawler work as universally first.

How often should you re-check?

Claude visibility is not static: models update, competitors publish, and your own site changes. A sensible cadence is a full re-check after every batch of fixes, then monthly. If you want it continuous, SearchScore’s Tracker puts real prompts to six live engines including Claude every week and counts exactly how often each one cites you, so you see movement instead of guessing.

What does a good result look like?

Benchmarks help contextualise readiness, not predict citations. The historical SAVI corpus had low average readiness and structure scores, and only a very small share reached SearchScore’s AI-Ready band. That band describes audit conditions; it is not a threshold where Claude citations become “likely”.

The pattern holds in the wild. Across 1,038 UK accountancy firms SearchScore audited, 97% let AI crawl them, yet only 18 (about 1 in 58) covered all five AI-readiness basics. Among those London firms the average GEO score was just 52.8/100. Being reachable is not the same as being recommended, and the gap between those two numbers is exactly what a Claude visibility check measures.

Related articles

• How Claude decides which websites to cite →
• Claude SEO guide: the full playbook →
• Why your website isn’t showing up in Claude →
• Free search visibility tools and checkers →

Sources & Further Reading

• Anthropic – Does Anthropic crawl data from the web? (ClaudeBot)
• SearchScore SAVI Report, Q3 2026 (Vol. 3, 1,000,000+ sites analysed)
• Academic research – GEO: Generative Engine Optimization (Aggarwal et al., arXiv)
• Schema.org – Organisation structured data reference