The mark
Agent Ready
A site that AI systems can actually read is doing something most of the web is not. The mark says so, in a form you can put in your own footer, and links back to the evidence.
744 of the 2,684 fully measured sites in the index qualify for a mark today. That is 28%.
No application, no account, no fee. If your site qualifies, the file already exists at a permanent URL and you can embed it right now.
The three levels
Two of them are awards. The third is a statement of fact, offered because some organisations want to show they are measured whatever the number says, and refusing them that would be its own kind of dishonesty.
Agent Ready
153 sites
Scores 90 or above. Allows the crawlers that answer questions, publishes machine-readable surfaces, and serves content an agent can actually read.
Agent Friendly
591 sites
Scores 75 to 89. Broadly open and legible to agents, with room left on the table.
Measured
1,940 sites
Measured and published, but scoring below 75. There is no award to embed at this level, and the site page lists what to fix in priority order.
How a site earns one
There is nothing to submit. Every domain in the nightly crawl is scored on the same rubric, and the mark follows the score.
- 1Be in the indexThe corpus is the most-visited domains on the web, plus anything anyone submits. Submitting takes one form and no account.
- 2Let the crawlers that answer questions read youForty-five of the hundred points. This is the one that moves scores, and for most sites it is a robots.txt edit rather than a project.
- 3Publish something machine-readableA sitemap declared in robots.txt, an llms.txt, licence terms. Twenty-five points, and the cheapest points on the board.
- 4Serve content a crawler can read without running JavaScriptThirty points, covering structured data, a dateline, an author, and text present in the server response.
- 5Score 75 or aboveThe mark is generated on the next nightly crawl. Nobody has to approve anything.
Measure your site now . Read the full rubric . Add a domain to the index
Take it
Three shapes, three syntaxes, one static file each. Enter your domain to get the exact snippet.
The mark links back to this site's page for your domain, so anyone can check the claim in one click. It regenerates nightly, which means it stays true and it can change.
The obvious questions
What does the mark actually certify?
That on the measurement date shown, an independent crawler requested this site the way an AI agent would and recorded the result. It certifies a measurement, not an endorsement, and every input is published.
Does it cost anything?
No. There is no account, no payment, no application and nothing to sign. The dataset is CC BY 4.0 and the mark is generated automatically for every domain that qualifies.
Can I lose it?
Yes, and that is the point. The mark regenerates from the nightly crawl. If the site starts blocking answer-surface crawlers or drops its structured data, the mark changes to match. A mark that could not be lost would not be worth showing.
What if the score is wrong?
Every line of it is arithmetic over archived evidence, and the evidence is downloadable. Open the site page, read which checks were awarded and why, and recompute it. If a check is genuinely misfiring, that is a bug worth reporting.
Will it slow my page down?
It is a static SVG of about one kilobyte served from a CDN with a long cache, marked lazy-loading in the snippet. It carries no script, no tracking and no cookie.
Why is there no mark for lower scores?
Because nobody would embed one, so offering it would be a pretence. Below grade B the useful thing is not a graphic, it is the prioritised list of what to change, which is on the site page and is usually shorter than people expect.
Currently at the top
Measured 2026-08-10. These change.
- 100railway.appAgent Ready
- 100crawlindex.orgAgent Ready
- 97crazyegg.comAgent Ready
Published by Fidget Labs BV. The mark, the rubric and the underlying data are free to use under CC BY 4.0. Download the dataset.