AIgentSphere

Free tools › llms.txt validator

llms.txt validator

Paste a file. You get a line per condition, and each line says where the rule comes from — Google’s agent-readiness audit, the llms.txt proposal, or our own measurements. Three separate authorities, never blended into one invented score.

Paste-in only, and here is why

There is no “fetch my URL” box on this page. A browser will not let this page read a file from your domain unless your server explicitly opts in to cross-origin reads, and almost none do — so a URL box would fail for most visitors and we would have to route it through a server of ours, which would mean receiving your file. Paste the text instead: open https://yourdomain/llms.txt, select all, paste here. Nothing leaves your browser.

check it

Your file

Nothing pasted yet.

the rules, and whose they are

What the audit actually checks

Worth knowing, because most advice about this file is stricter than the audit and quieter about which is which. What Google’s agent-readiness audit does with /llms.txt, from its own documentation and implementation: it fetches the file at the root path, and reports it as not following recommendations if there is no level-one heading anywhere in it, if it contains no Markdown links, or if it is very short. The package passes Google’s agent-readiness audit.

Three things the audit does not do, which matter if you are about to redesign a file around a rumour:

Verified against the audit’s documentation at developer.chrome.com and its implementation at release tag v13.4.1, 2026-07-20. The heading condition is stated in the audit’s own description; the minimum length and the link condition are traceable to the source file rather than to prose documentation, and this page says so rather than implying a document that does not exist.

three authorities

Where each check comes from

// how to read the verdicts
Google’s agent-readiness auditThe condition the audit applies. Failing one of these is the thing that makes the audit report your file as not following recommendations.
The llms.txt proposalThe format the specification describes. The audit is looser than the specification here; we follow the specification because it costs nothing to be correct twice.
Our own adviceNot in the audit and not in the specification. These come from what we measured about how agents read files and pages.

A tool that blends three authorities into one score out of a hundred is inventing the score. We would rather show you the three columns.

if it failed

Fixing it takes ten minutes

Build a clean one with the generator — it emits the same format our paid engine emits, and it self-checks against the same conditions as this page. Then use the per-platform instructions to get it served at exactly /llms.txt, which is the part that actually goes wrong on hosted platforms.

And keep the file in proportion. It is the compliance surface, not the mechanism: unpointed agents opened ours 0 times in 18 visits (Note 15), while the homepage fact block recovered 78% of fact checks against 30% in the control (Note 41). Pass the audit, then go and fix the page the agent actually reads.

Sources: MARKETING_NOTES_AGENT_VISIBILITY.md, Notes 15, 41.

free audit

See what AI sees on your site

A valid file is table stakes. What a model can recover from your homepage is the number that decides the enquiry.

One model fetches your homepage and answers buyer questions from it. You get the recovery score and the specific answers it could not find. That is the whole free tier: it shows you the problem, and the fix package is not included.