Question 1

What does the llms.txt validator check?

Accepted Answer

Three things most checkers skip. First, whether the file is genuinely retrievable: present at the root, served as text rather than as your site's HTML shell, and not blocked by your own robots.txt. Second, whether it matches the convention: a single H1, a summary blockquote, H2 sections, and links in the expected format with absolute URLs. Third, whether the links in it actually resolve, which we test live on a sample of up to twelve.

Question 2

Why does it matter if llms.txt is served as HTML?

Accepted Answer

It is the most common real failure and the hardest to spot. Plenty of sites have a catch-all route or a 404 page that answers with status 200, so the URL looks fine when you click it in a browser and what arrives is a page of HTML. A human sees a working link. A parser sees nothing it can use. We check the body rather than the Content-Type header, because the body is what has to be parsed.

Question 3

Does the validator check my links are live?

Accepted Answer

Yes, on a sample of up to twelve in document order, in parallel. A broken link inside llms.txt is worse than a missing one, because the file presents itself as an authoritative map of your site. We do not check every link, because on a file with several hundred that would take long enough to time out.

Question 4

What score should I be aiming for?

Accepted Answer

Anything with nothing in the fix column. The score is a summary of the findings rather than a thing worth optimising, and the findings are what to act on. For reference, the reference file published at llmstxt.org scores 100 on this check, and plenty of well-known documentation sites score in the seventies.

Question 5

Will fixing my llms.txt get me cited by AI?

Accepted Answer

On its own, no, and I would not sell it to you that way. llms.txt is a convention rather than a standard, no system is obliged to read it, and the evidence that any particular one does is thinner than the enthusiasm around it suggests. It is cheap to publish and cheap to get right. It is not the thing that decides whether a model names you, which comes down to entity clarity and third-party corroboration.

Question 6

Why does it flag placeholder text?

Accepted Answer

Because people publish drafts. Our own llms.txt generator deliberately writes REPLACE THIS against every description, on the grounds that an obvious placeholder is better than a plausible invented one. The failure mode is somebody publishing that draft unfinished, so the validator looks for our own marker along with TODO, FIXME and lorem ipsum.

Question 7

Do I need to give you an email address?

Accepted Answer

Not to run the check and not to copy the report. The validator runs and the copy button works with no email at all. We ask for one only if you want the report as a downloaded file, which is an honest trade rather than a wall.

Question 8

Does a long list of links count against me?

Accepted Answer

It raises a tidying flag past about a hundred. The point of llms.txt is to give a model a short, cheap map of what matters on your site. Past a certain size it stops being that and becomes a second sitemap, which already exists and is better at the job. The convention's answer is to keep llms.txt as the index and move the bulk into llms-full.txt.

Free llms.txt Validator

What we found

The fix list

What this does

The three checks worth having

The honest position on llms.txt

If you have not got one yet

What does AI say
about you?