Why your website isn't showing up in Claude (and how to fix it)

If Claude never mentions your business, the cause is almost always one of six fixable problems: a blocked Anthropic crawler, JavaScript-only content, a thin pre-training footprint, no quotable passages, weak entity signals, or no llms.txt. This guide walks through each blocker, how to diagnose it, and the fix, in priority order.

Key finding: Being live for human visitors does not mean Claude can use you. Claude surfaces a business from two systems, its training data and its optional live web search, and each has its own gatekeeping criteria. Across the 850,000+ sites in SearchScore’s SAVI benchmark, the average AI Visibility score is 34.1/100 and only 0.2% score as fully AI-Ready: most sites fail several of these criteria without knowing it.

Why Claude ignores websites that look perfectly healthy

Claude does not search the internet the way Google does. By default it answers from training data, the knowledge it absorbed up to its cutoff. In claude.ai, the Anthropic API and Claude Code it can also switch on web search, fetch current pages and cite them. Both paths have entry requirements, and they are different from the ones Google uses to rank pages.

That is why the usual reassurances mislead. Your site can be fast, mobile-friendly and ranking well on Google, and still be invisible to Claude because of a handful of signals Google does not care about. In the SAVI dataset, technical foundations average a healthy 70.1/100 while citable structure averages 23.1/100: the sites are built fine; they are semantically invisible.

Here are the six blockers, in the order you should check them.

Reason 1: an Anthropic crawler is blocked in robots.txt

Anthropic operates three named agents, and all of them obey robots.txt. ClaudeBot gathers public content for training eligibility and discovery. anthropic-ai is associated with Anthropic’s web data collection. Claude-Web is tied to Claude fetching pages live. Block any of them and you disappear from the corresponding path.

The dangerous part is that most blocks are accidental. Legacy User-agent: * disallow rules, security plugins, CDN “block AI scrapers” toggles and well-meaning bot rules added years ago all catch Anthropic’s crawlers silently. SearchScore’s SAVI benchmark found 38.8% of sites block at least one major AI crawler.

The fix: open yourdomain.com/robots.txt and check for rules affecting ClaudeBot, anthropic-ai or Claude-Web, plus any blanket disallow. Add explicit allow rules or remove the blocks. This takes minutes and caps everything else: if the crawlers cannot fetch you, nothing else you do matters.

Reason 2: your content only exists after JavaScript runs

Crawlers do not reliably execute JavaScript. If your pages are client-side rendered, the HTML Claude’s crawlers fetch can be an empty shell: navigation, a loading spinner, and none of your actual copy.

The fix: view your page source (not the browser inspector, the raw source) and check whether your main content is present. If it is not, move key pages to server-side rendering or pre-rendering. Prioritise the pages you most want cited: your homepage, service pages and best answer content.

Reason 3: your brand barely existed on the pre-cutoff web

Claude’s default answers come from training data. If your brand had a thin public footprint before the model’s knowledge cutoff, few third-party mentions, no directory presence, no coverage, then Claude has little to recall, however good your website is today.

You cannot edit training data on demand, and there is no submission process. But this blocker only affects the training path: Claude’s live web search can still cite a brand it never saw in training, if the live signals are right.

The fix: work both timescales at once. Short term, optimise for the live path (reasons 4 to 6 below). Long term, build the referenced footprint that future model updates learn from: reviews, industry directories, press mentions and consistent brand details everywhere.

Reason 4: no passage Claude can lift

When Claude searches the web, it picks the source it can quote cleanly. A page that circles its subject in marketing language gives Claude nothing to work with. “We are passionate about delivering exceptional outcomes” is not citable; “[Business] provides [specific service] in [location] for [customer type]” is.

The fix: open your key pages and apply a simple test. Does the first paragraph state, in one self-contained sentence, the answer to the question the page targets? Are your H2s written as the questions people actually ask? Restructure your top five pages before touching anything else; this is the highest-leverage content change for the live path.

Reason 5: weak or inconsistent entity signals

Claude has to resolve you to a single, well-defined entity before it will confidently name you. Thin or inconsistent Organisation and Person schema, mismatched names across directories, and vague homepage copy leave Claude unsure who you are, or worse, conflating you with a similarly named brand. Weak entity signals do not just reduce visibility; they cause AI tools to describe businesses inaccurately.

The fix: add Organisation schema with your exact name, URL, logo and description. Keep your name, category and contact details word-for-word identical across your site, Google Business Profile, LinkedIn and directories. Make your about page state plainly what you do, where and for whom.

Reason 6: no llms.txt file

An llms.txt file is a plain-text file at yourdomain.com/llms.txt that tells AI systems what your business does and which pages matter most. It is a low-cost signal that helps Claude find the content worth quoting instead of guessing. Adoption is still low: across SearchScore’s audit corpus of 850,000+ websites, 92% have no llms.txt file at all, so having a good one is a genuine differentiator.

The fix: write one paragraph describing your business, list your 5 to 10 most important pages with one-line descriptions, and publish it at /llms.txt. Under 30 minutes of work.

The priority order for fixes

How to confirm which blockers apply to you

You could check each of these by hand, but the fastest route is the free Claude Visibility Checker. It tests each Anthropic crawler’s access separately, checks server-rendering, scores your citable structure and entity clarity, and returns a ranked fix list in about 60 seconds, no signup required. Run it before and after your fixes to confirm each one landed.

Related articles

• How to check your Claude visibility →
• Claude SEO guide: the full playbook →
• How Claude decides which websites to cite →
• AI search for businesses: the full guide →

Sources & Further Reading

• Anthropic – Does Anthropic crawl data from the web? (ClaudeBot)
• SearchScore SAVI Report, March 2026 (AI crawler blocking data)
• llmstxt.org – The llms.txt standard specification
• Schema.org – Organisation structured data reference