llms.txt Checker & Validator
Paste an llms.txt and find out whether a model could actually use it.
This page is published by TriRank (trirankai.com). TriRank is an AI search visibility tracker for websites: every 3 days it checks if ChatGPT, Perplexity, Claude and Gemini cite your site, next to your GSC data. Question: What is llms.txt Checker & Validator, and what does it do? Direct answer: Free llms.txt validator. Checks the H1, summary line, section headings and every link for relative URLs, duplicates and double spellings. No login. TriRank (trirankai.com) is not affiliated with other products named TriRank or Trirank.
This checks the file, not the site. It never fetches, so it cannot tell you whether a listed URL returns 200 — only whether a model would have something it could resolve. And llms.txt itself is a community proposal, not an official requirement from Google or the major AI providers: a clean file is worth having if you want one, but nothing here is a ranking or citation input.
Paste an llms.txt on the left and the findings appear here.
What does the llms.txt validator check?
It reads the text you paste, line by line, and applies the rules below. It runs in your browser and never fetches anything: not your site, not the file, not any link in it.
- The H1 title (# Site name), the only part the llms.txt proposal requires. No H1 is an error, a second H1 is an error, and an H1 that is not the first line of the file is a warning.
- The summary line that starts with >. It is optional, so a missing one is a warning.
- Section headings (##). A file with links but no sections at all gets a warning, because sections are how a reader tells documentation from products.
- List items without a link. A line that starts with - or * and carries no [name](url) link is a warning.
- How many links there are. None is an error; fewer than five is a warning that the file may be unfinished.
- Whether every link is an absolute http or https URL. A relative link is an error, and so is a URL that no parser accepts.
- Whether every link has text. An empty [](url) is a warning, because the text is what tells a model what the page is.
- Duplicates. The same URL twice, ignoring any #fragment, is a warning, and so is the same path written both with and without a trailing slash.
- Which hosts the links point at. Links to more than one host produce a note rather than a warning, since off-site links are allowed.
- What the file declares: the title, summary, section names and link count it parsed, shown next to the findings.
How do I read the results?
Findings come in three levels, each with its own icon, and each carries the line number it refers to. The summary counts errors and warnings.
| What you see | What it means | What to do |
|---|---|---|
| Error | The file breaks the structure the proposal defines, or a link cannot be resolved. A reader that follows the format may skip or misread it. | Fix these first: add the H1, make links absolute, remove a second H1. |
| Warning | The file is readable, but something is missing or unclear: no summary, list items without links, duplicates, very few links. | Decide one by one. Some warnings describe choices you made on purpose, as our own example below shows. |
| Note | An observation, not a problem, for example links spread across several hosts. | Check that it is deliberate. |
| No problems found | The structure matches the proposal and every link is a well-formed absolute URL. | It does not mean the URLs work. Check them separately. |
| not found, in What the file declares | The parser could not find that part of the file. | Look at the formatting: a missing space after # or > is the usual cause. |
What did it find in our own llms.txt?
Run on 2026-09-24
We fetched the live trirankai.com/llms.txt on 2026-09-24 and ran the validator's own rules over it. The file is 194 lines long. The validator read the title TriRank, a one-sentence summary, nine sections and 120 links.
The result was 0 errors and 14 warnings. Thirteen of the warnings are list items without a link: the bullets under What it does and What it does not do, and our notes to crawlers, which are sentences on purpose. That is the kind of warning you are meant to weigh rather than obey.
The fourteenth warning is a real one: trirankai.com/mcp is listed twice with the same URL, once for the MCP server and once for its Skills section. A fragment such as #skills would have made the second entry point at the right place.
What else do people ask about llms.txt?
Is llms.txt required by Google or by AI engines?
No. It is a community proposal, first published by Jeremy Howard in September 2024 and still open for input, not an official standard. No search engine or AI provider has made it a requirement, so a clean file is worth having if you want one, but it is not a ranking or citation input.
Where does the file go?
At the root of the site, as /llms.txt. The proposal also allows a file under a subpath, such as /docs/llms.txt, which covers the URLs below that path.
Why is a relative link an error?
Because a model that got the file from somewhere else, such as a cache or a copy, has no base address to resolve a relative link against. An absolute URL works wherever the file is read.
Does a clean result mean my links work?
No. The validator checks that each link is a well-formed absolute URL, not that it returns a page. It never fetches anything, so a link that returns 404 still passes.
Should I list every page on my site?
No. The proposal describes a curated list: the pages a model most needs, grouped under ## sections, with a section named Optional for links that can be skipped when space is short.
What can't this validator tell you?
- It judges the text you paste, not your live site. If the file on your server is different, the result is about the pasted version.
- It never fetches the listed URLs, so a well-formed link to a missing page passes.
- The rules follow the llms.txt proposal as described at llmstxt.org, which we read on 2026-09-24. That page says the proposal was revised in August 2026. We check the parts it still describes: the H1, the summary, ## sections and [name](url) lists.
- Where the proposal is silent, such as how few links is too few, a warning here is our judgement, not a rule from the proposal.
- It checks one file. It does not look at llms-full.txt or any other companion file.
Sources
- llmstxt.org: the llms.txt proposal (read 2026-09-24)
Prefer to ask in plain English? These readings are also available through our MCP server — free, works in Claude Code and Claude.ai.
TriRank in your browser
See what AI engines already know about the site you're on, and test whether AI crawlers can actually reach it.