Does this checker measure AI rankings?

No. It records observable machine-readable evidence. It cannot guarantee indexing, citation, ranking, trust or agent use.

Does H/M store the website I scan?

No target URL or fetched page body is stored. H/M keeps only aggregate scan counts by channel, outcome and country, plus a rotating anonymous quota hash.

Is llms.txt an official web standard?

No. The scanner labels llms.txt as a community proposal and reports it separately from normative standards such as RFC 9309, RFC 9116 and RFC 9727.

H/M · AGENT SURFACE SCANNER

Can an AI discover, understand and use your website?

Inspect the evidence an AI crawler or agent can actually observe. No guessed ranking, no citation promise, no hidden upload.

Live HTTP evidence Not a ranking score Private networks blocked

Four evidence layers: access, retrieval, action and trust.

01 / LIVE AUDIT

Evidence surface

Enter a public domain to begin.

01Accesscrawler policy

02Retrievalcontent signals

03ActionAPIs and agents

04Trustpolicy and integrity

One URL, four evidence layers

The scanner requests only the homepage and common public machine artifacts. Response bodies are analyzed in memory and are never returned or stored.

02 / METHOD

Evidence before advice

Each result names the requested public artifact, what was observed and which specification or convention it relates to. Missing evidence is not converted into a fake probability.

Can crawlers enter?

HTTPS, homepage reachability, robots.txt, named AI user agents and Content Signals.

Can systems retrieve context?

Sitemap, llms.txt, metadata, JSON-LD and optional Markdown negotiation.

Can an agent act?

OpenAPI, API Catalog, MCP discovery, A2A Agent Card and payment discovery.

Can claims be checked?

security.txt, policy links, cache validators and evidence digests in the paid package.

03 / SOURCES

Standards and declared conventions stay separate

OpenAI Crawlers OAI-SearchBot, GPTBot and ChatGPT-User have different purposes. Cloudflare Content Signals Search, AI input, training and content-use preferences. Model Context Protocol Registry metadata and capability discovery are reported without a trust score. A2A Agent Discovery The well-known Agent Card describes identity, endpoint, skills and security.

04 / LIMITS

What this audit does not prove

It does not prove that ChatGPT, Claude, Gemini or another system indexed the site.

It does not estimate citation, ranking or traffic probability.

A self-declared MCP or A2A identity is not cryptographic identity proof.

It records one moment in time; CDN, WAF and origin behavior can change.