EditionsFour reports, on a schedule, from measurements anyone can check.
Each edition has a live page that updates as the measurements do, an Atom feed, signed JSON, and a written post for each issue. Figures come from the same records the API serves; nothing in an edition is estimated.
Monthly
What sites serve AI crawlers across the open dataset and the weekly census of the most-visited sites: refusals, machine files, what a crawler actually receives.
Latest: 139,788 homepages in four days: what AI crawlers get
First: The state of AI visibility, September 2026: what 1,364 scans say a crawler actually receives
The live pageAtom feedJSON
Monthly, with weekly signed editions
What each named AI and search agent did on this site, from requests verified against its operator's published ranges: robots.txt, forged names, busiest minute.
Latest: Agent conduct, October 2026: 34,676 verified crawler requests, and what they did with robots.txt
The live pageAtom feedJSON
Monthly, sealed after the month closes
Whether the remote MCP servers in the official registry answer, how fast, and how often their tool lists change. No tool is ever called.
Latest: MCP reliability, October 2026: 6,000 remote MCP servers, 39,484 handshakes
The live pageAtom feedJSON
Monthly, sealed after the month closes
Of the agent requests connected sites verified, the share that arrived with a CrawlCheck receipt the site accepted.
Latest: Share of preflight, October 2026: 0 of 5,643 verified agent requests carried a receipt
The live pageAtom feedJSON
How the editions are made
- Every figure links to the record it came from; the records are signed and verify offline with /crawlcheck-verify.mjs.
- Monthly editions seal after the month closes (UTC) and are never edited afterwards; a correction is a new, dated note.
- No site under investigation is named, ever. Agents and platforms are named only where the record names them.
All findings and guides: /blog · the full Atom feed · the open data: /data.