State of the agent-readable web

How we count.
And what we don't claim.

The census is a measurement of adoption and access — which domains publish files an AI agent can read, and which bot directives they name or block. It is deliberately not a measurement of citations, recommendations or visibility inside AI systems. This page is the single reference for how every edition was collected.

Universe

What we look at.

Each edition crawls the top 100,000 domains of the Tranco list, taken on the day of the crawl. Eight surfaces per apex domain:

A surface is added, never substituted, and carries the date it entered. When the ARD spec moved its canonical path in August 2026, the obvious move was to swap one column for the other. We did not: a swap would have broken the month-over-month series at the exact point where the count of ARD publishers is single digits, and a reader comparing two editions would have seen a collapse that was ours, not the web's. Both paths are counted in parallel. A column that starts mid-series reads zero for the editions before it, and that zero means not measured, not nobody published — the dates above are how you tell the two apart.

Subdomains are out of scope by design: docs.example.com publishing an llms.txt does not count for example.com. The census measures apex-domain adoption, the same bar for everyone.

Collection

How we crawl.

Plain GET requests, no rendering, against each path on each apex domain, paced at ~4 requests/second globally, honoring robots.txt, identified as DesvelaBot/0.1 (bot policy).

Coverage

When we publish a number.

Only editions with complete or explicitly declared coverage are published. A crawl that does not finish is not silently reported as full: the generator refuses to emit a census unless at least 95% of the universe was crawled within the last 3 days, and a partial pass must declare its own denominator. An incomplete internal pass that never became a public edition does not appear in the archive, in downloads or in the editions list.

Each edition states, in its manifest, the number of domains checked, the universe size, when the data was collected, when it was published and when (if ever) it was corrected.

Honesty rules

What these numbers do not claim.

When a published figure turns out to be wrong, the number that went out stays and a dated correction sits next to it. Quietly rewriting a number somebody may have already cited is worse than having been wrong in public.
Editions & data

Where everything lives.

Each edition keeps its own URL, so a number cited last month is still there to check. Every published edition offers a downloadable JSON manifest (provenance + metrics) and a CSV of the tier aggregates.

Latest edition →  ·  July 2026  ·  August 2026