You decided who may train. Did you check who may crawl?
AI answers are built from pages bots can fetch. Host defaults and bot fights can block GPTBot and friends even when robots.txt looks open. We measured this on real small sites: in a 60-site corpus, robots-level search-AI disallow was essentially absent, while about 18% still turned AI user-agents away at the edge. Details in our crawler study and corpus notes.
Free whole-site scan · AI crawlers by name · no email on the diagnosis
Access first. Then structure.
Named AI crawlers
GPTBot and related probes, not a single vague "bots" toggle.
Robots and edge behavior we can see
When the edge rejects an AI user-agent, the report says so in plain language.
Pages worth citing
Answer blocks, schema, depth. Access without quotable pages is half a strategy.
We do not sell "get cited tomorrow." We sell an honest map of what is broken. Related tools: AI crawler access checker, llms.txt generator.
What this is not.
We are not a full live citation tracker for ChatGPT answers (that is a different, expensive product class). We measure readiness and access on your site: the half you control. Method notes stay public in the learn library and studies.