Free tool · no signup · 10-second scan
Can AI actually read your website?
ChatGPT, Claude, and Perplexity now answer your customers' questions about you — whether you like the answers or not. Scan any domain and get its AI Legibility Score: five live checks that decide whether AI cites your real docs or makes things up.
The five signals we check
Every check runs live against your domain, from the outside — exactly the way an AI crawler or agent sees you. Nothing is stored.
llms.txt published
A plain-text map at yoursite.com/llms.txt that tells AI assistants where your canonical content lives.
Why it matters: Assistants fetch it before answering questions about you. Without it, they answer from stale training data — or from whoever ranks above you.
AI crawlers allowed
robots.txt rules for GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, and Google-Extended (Gemini).
Why it matters: A single Disallow line makes you invisible to an entire AI ecosystem. Many sites block these bots by accident via copy-pasted robots files or CDN defaults.
MCP endpoint live
A Model Context Protocol server that AI agents can query in real time — the API layer of the agent web.
Why it matters: Agents that can query you directly cite your live docs instead of guessing. Almost nobody has this yet, which is exactly why it wins.
Sitemap discoverable
A sitemap.xml declared in robots.txt or served at the standard path.
Why it matters: It is how crawlers — search and AI alike — discover pages beyond your homepage. No sitemap means partial coverage of what you publish.
Machine-readable metadata
A homepage <title>, meta description, and ideally JSON-LD structured data.
Why it matters: The minimum viable self-description. If a machine cannot tell what your site is, it will not recommend it to anyone.
Score below 100?
HelpShelf publishes every one of these signals for you — generated from your real docs, kept fresh automatically — with one command: npx helpshelf
AI legibility, explained
- What is an AI legibility score?
- It measures whether AI systems — ChatGPT, Claude, Perplexity, Gemini, and the agents built on them — can read, understand, and cite your website. The score checks five outside-in signals: an llms.txt file, robots.txt access for AI crawlers, a live MCP endpoint, a discoverable sitemap, and machine-readable homepage metadata. It is the AI-era equivalent of a technical SEO audit.
- How is the score calculated?
- Each of the five signals is checked live against your domain and marked pass, fail, or unknown. Your score is the percentage of known signals that pass — if a check cannot complete (a network timeout, for example), it drops out of the calculation instead of counting against you. 80+ is AI-legible, 40–79 is partially legible, below 40 means AI mostly cannot see you.
- Why does it matter if ChatGPT can read my site?
- A growing share of buying research now happens inside AI assistants instead of search engines. When someone asks ChatGPT or Perplexity about your product category, the sites those assistants can actually read are the ones that get cited and recommended. If your site is illegible, the answer comes from stale training data — or from a competitor who did the work.
- Am I blocking AI crawlers without knowing it?
- Very possibly. Many robots.txt templates, CDN bot-protection defaults (including some Cloudflare managed rules), and copy-pasted configs block GPTBot, ClaudeBot, or PerplexityBot wholesale. This scan reads your live robots.txt and tells you exactly which AI crawlers are blocked, so you can decide deliberately instead of by accident.
- What is an MCP endpoint and do I really need one?
- MCP (Model Context Protocol) is the open standard that lets AI agents query a site directly — think of it as an API designed for AI assistants. It is early: most websites do not have one, which is precisely why having one makes you stand out to the agents that check. HelpShelf hosts an MCP server for every site automatically.
- How do I improve my score?
- Each failed check comes with a specific fix: publish an llms.txt with our free generator, unblock AI crawlers in robots.txt, add a sitemap, and fill in your homepage metadata. Or run npx helpshelf — it publishes llms.txt, llms-full.txt, a hosted MCP endpoint, and an AI-welcoming help center for you; re-run it when your source changes.
Related: build your llms.txt with our free generator, read the answer engine optimization guide, or see how AI agents read your product.