Free llms.txt Validator
Check your llms.txt for the things that quietly break it: missing sections, links that no longer resolve, and a file that has grown into a copy of the site.
What a valid file looks like
The format is small on purpose. One title, an optional summary, then sections of links with a short note on each. These three llms.txt examples cover most of what a site needs.
A small business site
# Leeds Bakery Co > Family-run bakery in Leeds, wholesale and retail. ## Products - [Wholesale ordering](https://example.com/wholesale): minimums, lead times and delivery days - [Allergen information](https://example.com/allergens): full ingredient breakdown per product ## About - [Contact and opening hours](https://example.com/contact): address, phone and hours
A documentation site
# Chartsmith > Open-source charting library for the web. ## Docs - [Quick start](https://example.com/docs/start): install and render a first chart - [API reference](https://example.com/docs/api): every option, with defaults - [Migration guide](https://example.com/docs/v3): what changed in version 3 ## Optional - [Changelog](https://example.com/changelog): release history
A SaaS product
# Ledgerly > Invoicing and expense tracking for freelancers. ## Product - [Features](https://example.com/features): what the product does - [Pricing](https://example.com/pricing): plans and limits ## Support - [Help centre](https://example.com/help): setup, billing and troubleshooting
The ## Optional heading has a specific meaning in the proposal: it marks links an assistant may skip when it is short of room. Everything else is treated as worth reading.
What this tool checks
Missing sections. A file needs exactly one # Title line, and it should carry a > summary and at least one ## section grouping its links. A heading with nothing under it is reported too, since an empty section tells a crawler nothing.
Broken links. Up to 20 of the links in the file are fetched and the ones that do not return a page are listed with their status. Relative links are flagged separately: an assistant reading the file has no page to resolve them against, so they need the full address.
File size. Reported in bytes, with a warning past 100KB. The file is meant to be an index of the pages worth quoting, not a mirror of the site.
The most common failure is none of those three: the site answers /llms.txt with an HTML page, because a catch-all route handles the request instead of a real file. That returns a 200, so anything checking only the status code calls it present.
How to fix what it reports
The file came back as HTML. Your host is answering /llms.txt with a page rather than a file. On a single-page app or a framework with a catch-all route, the fallback swallows the request before it reaches anything static. Put the file where the host serves static assets from — public/, static/ or the published output folder — and add an exception ahead of the catch-all if the framework needs one. Then reload the address in a private window and confirm you see plain text rather than your site.
Relative links. An assistant reading the file has no page to resolve /pricing against, because it is reading a text file rather than browsing your site. Write every link as a full address, starting https://. This is the most common thing a hand-written file gets wrong.
An empty section. A ## heading with no links under it tells a crawler nothing and makes the file look half-finished. Either put a link under it or delete the heading. It is better to ship three sections that are full than six where half are placeholders.
A link that does not resolve. Usually a page that moved after the file was written, which is the failure mode of any hand-maintained index. Fix the address, or drop the entry if the page is gone. A file pointing at 404s is worse than a shorter one, because it spends the reader's patience on nothing.
The file is too large. Past about 100KB you are no longer writing an index, you are mirroring the site. Cut it back to the pages you would actually want quoted in an answer. Most sites need far fewer entries than they first assume — the useful version is closer to twenty links than two hundred.
What the llms.txt validator does not check
Worth being clear about the edges, because a green result is easy to over-read. This tool checks that the file is structurally sound and that its links resolve. It does not know whether you picked the right pages, and that is the part that decides whether the file is any use. A perfectly valid file listing your least important pages will pass every check here and do nothing for you.
It also does not fetch every link. Up to 20 are checked, which covers most files completely and gives a representative read of a longer one. If yours is much bigger than that, the size warning above is the more useful signal anyway.
And it cannot tell you whether an assistant read the file, because no engine reports that. Nobody can tell you that today. What you can measure is the outcome rather than the mechanism: whether the assistants name and cite you at all. That is what the free AI visibility checker is for, and it needs no account either.
Frequently asked questions
What does this validator check?
Three things. Missing sections: whether the file has the single title line it requires, a summary, and at least one section grouping its links. Broken links: it fetches up to 20 of the links the file names and reports the ones that do not return a page, plus any that are relative rather than absolute. And file size, since llms.txt is meant to be a short index rather than a copy of the site.
Where should llms.txt live?
At the root of your domain, as https://yoursite.com/llms.txt, served as plain text. The most common failure this tool finds is a site answering that address with an HTML page, because a catch-all route or a single-page-app fallback handles the request instead of a real file. That returns a 200, so anything checking only the status code reports the file as present when a crawler would find a web page.
Is llms.txt a standard?
It is a proposal rather than a ratified standard, and no engine has committed to reading it. It costs very little to publish, it is easy to keep accurate, and a well-made one doubles as a map of the pages you would want quoted. Treat it as cheap insurance rather than a ranking lever.
Do I need an account to use this?
No. This validator runs with no account, no card and no credits, and it works signed out. Paste a domain and it fetches the live file and checks it.
I do not have an llms.txt yet. What now?
Use the llms.txt Generator, which reads your sitemap and page titles and writes a file you can publish as it stands, then bring it back here to check it once it is live. The generator page also carries three short examples of what a finished file looks like.
Does a valid llms.txt mean AI will cite me?
No, and anyone promising otherwise is selling something. llms.txt is a proposal that no engine has committed to reading, so a valid file is a thing you have made available rather than a thing that has been consumed. What it does reliably is remove an excuse: if an assistant does read it, the file is correct, points at pages that exist, and describes them in your words rather than in a guess assembled from your navigation. Judge the outcome separately, by checking whether the engines actually name you.
How often should I re-check it?
Whenever the pages it names change, which in practice means after a site restructure, a URL migration, or a batch of new pages you would want quoted. The file itself does not rot, but the links in it do: the usual way a good llms.txt goes bad is that somebody moves a page six months later and nothing points at the file to say so. If you publish often, a check every quarter is enough to catch that before it matters.
Related
llms.txt Generator writes the file for you from your sitemap. llms.txt Explained covers what the file is for and how to write one by hand. All free tools.
Track llms.txt, schema and AI visibility across every page, on a schedule.