How it works

AIScan.site grades how prepared your website is for AI agents and LLM crawlers — ChatGPT, Claude, Perplexity, Gemini, and the long tail of agent runtimes that are starting to read, reason about, and act on the web on your behalf.

What we fetch

A scan takes about 20 seconds and is read-only — we never write to your site. From a single URL, we make a small number of polite HTTP requests:

· Your homepage (and follow redirects)
· /robots.txt and any Sitemap: it declares
· /sitemap.xml or /sitemap_index.xml
· /llms.txt (the emerging llmstxt.org standard)
· The homepage again with Accept: text/markdown
· Seven /.well-known/* endpoints (API catalog, MCP server card, agent skills, OAuth metadata, UCP, Web Bot Auth keys)

User-agent: AIScanBot/1.0. Body responses are capped at 512 KB. Everything else is ignored.

Platform detection first

Before scoring, we identify the platform — WordPress, Shopify, Lovable, Replit, Bolt, v0, Framer, Wix, Webflow, or custom. We use server headers, host patterns, meta generators, and asset URLs. The detected platform reweights the rubric: a WordPress site is graded harder on content and discoverability, a Shopify store on commerce, an AI-builder app on capabilities.

Five dimensions

Discoverability

Can agents find your pages? robots.txt, sitemaps, and discovery headers.

Content

Can agents read your content cleanly? Markdown content negotiation, llms.txt, structured HTML.

Bot Access

Are AI crawlers explicitly addressed? Content Signals, per-bot rules, Web Bot Auth keys.

Capabilities

Can agents do things? API catalogs, MCP server cards, agent skills, OAuth metadata.

Commerce

Can agents transact? Universal Commerce Protocol, x402 payment signals. Only scored for e-commerce sites.

How a check is scored

Each check has a weight in points and produces one of three outcomes:

Pass — full points. The standard is implemented well enough that an agent can rely on it.
Partial — half-ish points. Something exists but isn't quite right.
Fail — zero points, with a one-line fix you can hand to ChatGPT or Claude.

Dimension scores are weighted averages within the dimension. The overall score is the weighted average across applicable dimensions — Commerce is excluded if you're not an e-commerce site, for example.

Six levels

Lv 5Agent-Native90 – 100

Lv 4Interactive70 – 89

Lv 3Governed50 – 69

Lv 2Readable30 – 49

Lv 1Discoverable10 – 29

Lv 0Invisible0 – 9

What we don't do

· No login, no JavaScript execution, no crawl past the homepage.
· No data is stored on our servers — the share link encodes the full result in the URL.
· No SEO ranking signals — this is about agent readiness, not Google rank.

Scan a site