llms.txt Checker
Looks for the llms.txt file at your domain root and, if it finds one, checks the structure: are the title, summary, sections and links in the expected shape?
- Sign-in requiredMembership is free
- Open on the Pro plan100 runs a day
- Saved to your historyReachable from your account
What does this tool look at?
Can ChatGPT, Perplexity and Google AI Overviews read you and cite you? Readiness checks for the new generation of search.
- Both llms.txt and llms-full.txt looked up
- Title, summary and section structure checked
- Link count and format verified
- A starter draft suggested if the file is missing
What this tool measures
llms.txt is a plain-text file telling language models what your site does and which pages matter. Publishing one is easy; publishing a malformed one is easier still: a missing title line, relative links, or a server that marks the file as a download.
The tool fetches llms.txt from your domain root and audits eight things: whether the file is reachable and how large it is, the H1 title on the first line, the summary line beneath it, section headings, link count, whether links are absolute, the content type the server sends, and whether llms-full.txt exists.
When the file is missing the audit scores zero and, instead of an empty warning, prints a starter template already filled with your domain. Copying and completing that is faster than writing one from scratch.
How to use it
- 1
Enter the domain
The file is looked for at the root. If a subdomain publishes separate content, such as docs.example.com, audit it separately; the root file does not cover it.
- 2
Read the checks top to bottom
Access and title are foundations; links and content type are refinements. Fixing the lower items while an upper one fails gains nothing.
- 3
Review the file contents
The result prints the first 6,000 characters of the file. If what the server actually sends differs from what you believe you wrote, you see it here.
- 4
Fill the gaps with the generator
The llms.txt Generator builds the same structure in the correct order. Regenerating and replacing usually produces a cleaner file than patching the old one.
- 5
Check AI crawler access as well
A perfect file is unreadable if robots.txt blocks AI crawlers. Verify both together with the AI Crawler Access Checker.
When to use it
- Confirming that a freshly published llms.txt is genuinely reachable.
- After a hosting or server change, to see whether root-level files survived the move.
- To study how a competitor or an industry leader structured their file.
- When AI answers describe you incorrectly, to see whether the summary line is the cause.
Reading the result
H1 title failure
The file must open with “# Site Name”. That line is where a model anchors who it is talking about. A blank line or a comment above it hides the title from the check.
Summary block warning
Directly under the title a one-sentence summary starting with “> ” is expected. That sentence is what most models reuse when describing you; leave it out and the model writes its own from page text.
Relative link warning
Links written as /services may be unresolvable, because a model can read the file detached from its original context. Make every link a full https address.
Content type warning
The server should send text/plain or text/markdown. A type such as application/octet-stream means the file is marked as a download, and some clients never read it.
The llms-full.txt note
This line is informational, not a deficiency. A full-text version makes sense for documentation sites; a small corporate site loses nothing by skipping it.
Questions about the llms.txt Checker
The score averages eight checks. Even a reachable file loses points for missing section headings, relative links or a wrong content type. The note beside each line names the gap.
No such measurement exists, and the file is not an official standard. Models can read your site without it. What the file buys you is not leaving your own description to chance.
Whenever a product, service or key page changes. Reviewing twice a year suits most sites, and the part that matters most is leaving no dead links behind.
The audit wants at least one and fails with none. In practice five to twenty works well: the pages a model should open when describing you, not the whole site.
Use it alongside these
The other tools in the same category. Run them together and you see the whole picture.
The tool found the problem. Now for the fix
We run technical SEO, content and AI visibility work end to end. Let us review your site and come back with a clear list of what to do.
Monday – Saturday 09:00 – 18:00