Agent readiness statistics
How many sites have agent readiness signals, in four samples, and why the numbers differ.
In short
Adoption of agent readiness signals is low and depends on the sample: markdown negotiation reaches 9 percent of 8,822 domains in this site's crawl and 3.9 percent in Cloudflare Radar's scan of 200,000. Content Signals sit at 7.4 percent in the crawl and 4 percent in Radar.
The same signals in two samples
| Signal | Shared crawl, 4 Nov 2026 | Cloudflare Radar, April 2026 |
|---|---|---|
| Serves a robots.txt | 79.2% of 10,886 domains that answered | 78% of the filtered top 200,000 1 |
| Declares Content Signals | 7.4% of 8,624 domains with a robots.txt | 4% of sites |
| Answers Accept: text/markdown with markdown | 9% of 8,822 home pages | 3.9% of sites |
| MCP server card or API catalog | 0.9% serve /.well-known/mcp.json (103 of 10,886) | fewer than 15 sites together in the dataset |
The shared crawl read 13,071 domains on 4 November 2026 with a plain-fetch client: the Tranco top 10,000 plus Ora's leaderboard universe, sites that someone scanned with Ora. Radar scanned the 200,000 most visited domains after dropping categories where agent readiness is unimportant 1. The data page has every row of the crawl.
Protocol adoption in agent-ready.dev's corpus
| Protocol surface | Sites | Share |
|---|---|---|
| API catalog (RFC 9727) | 330 | 13% |
| MCP server card | 280 | 11% |
| Agent Skills discovery | 270 | 10% |
| A2A agent card | 259 | 10% |
| Web Bot Auth directory | 182 | 7% |
| agent-permissions.json | 165 | 6% |
| UCP profile | 165 | 6% |
| agents.json | 164 | 6% |
Source: agent-ready.dev's State of Agent Readability page 2. It is a sample of sites whose owners chose to scan, so rates run ahead of the web at large.
Scores in Ora's corpus
Ora's stats on 5 November 2026 give an average score of 44 over 132,085 ranked companies, with 784 at A or better, a share of 0.6 percent 3. The breakdown by category is on the industry page.
What the crawl says about llms.txt
24.6% of 10,598 domains served /llms.txt in the crawl, far above the share that serve a markdown alternate on the home page (2.5%). A file that exists is not the same as a file an agent reads: Ora's lab found that 86 percent of llms.txt fetches in its traced runs arrived through a link, not a guessed path (vendor claim, 19 October 2026).
Sources
- blog.cloudflare.com/agent-readiness. https://blog.cloudflare.com/agent-readiness/, read 2026-11-05.
- agent-ready.dev/state-of-agent-readability. https://agent-ready.dev/state-of-agent-readability, read 2026-11-05.
- ora.ai/api/stats. https://ora.ai/api/stats, read 2026-11-05.
Last reviewed 2026-11-05