Check your llms.txt
An llms.txt validator reads the llms.txt file at the root of a site and checks it against the llmstxt.org format: an H1 with the site name first, a blockquote summary, and H2 sections listing pages as markdown links. Enter your store address to see each error and warning by line number, and whether the pages it links to open.
The format, in one example
A file that passes every check. The first line is the only one the format requires; the rest is what makes the file useful to an assistant.
# Harbour Supply Co
> Waxed cotton bags and outdoor clothing, made in Cornwall and shipped across the UK and EU.
## Products
- [Harbour Duffel](https://harbour.example/products/duffel): 40 litre waxed cotton holdall
- [Field Jacket](https://harbour.example/products/field-jacket)
## Policies
- [Returns](https://harbour.example/policies/returns): 30 days, free in the UK
## Optional
- [Our story](https://harbour.example/pages/about)What we check
An error is something the format requires and the file breaks: no H1 on the first line, a second H1, a link that does not parse, or a web page served where the file should be. A warning is something the format recommends, or something that makes the file less useful: no summary, a section with no links, a relative link, a link listed twice. #
We also read /llms-full.txt if you publish one, and follow up to 5 of the links on your own site that your llms.txt lists. A link that answers 404 or 410 is broken; one that redirects works, and is still worth pointing at where it lands.
No llms.txt yet?
The llms.txt generator reads your sitemap and writes a starting point from your own pages. Publish it, then check it here. Agent view shows how one product page reads to an AI agent, and the robots.txt checker shows which agents may read it at all.
Frequently asked questions
What is a valid llms.txt?
One that follows the format proposed at llmstxt.org. It is a markdown file at /llms.txt that starts with an H1 naming the site. A blockquote with a short summary comes next, then any notes, then H2 sections. Each section is a list of markdown links, one per page, written as [Page name](https://...), with an optional colon and a note after it. The H1 is the only part the proposal requires.
What is the difference between llms.txt and llms-full.txt?
llms.txt is an index: it names your site and lists the pages worth reading, with a link to each. llms-full.txt is the content of those pages in one file, so an assistant can read it without following the links. The proposal treats both as optional, and llms.txt is the one an assistant looks for first.
Why does my llms.txt show as a web page?
Your server is answering /llms.txt with HTML. That usually means the address is not a file at all, and a catch-all page (often your home page or a not found page with a 200 status) is being served in its place. Publish the file itself at that address. On Shopify, check that an app or theme setting is not overriding the default file.
Do AI agents actually read llms.txt?
Some do and many do not yet. Evidence that it changes what an assistant recommends is weak compared with a product feed or structured data on your product pages, which is why a scan reports a missing llms.txt as the lowest severity finding it has. It is quick to get right, and getting it right means checking the format, because a file an assistant cannot parse is no better than none.
What do you send to my store?
robots.txt first, then /llms.txt and /llms-full.txt, then up to 5 of the links on your own site that your llms.txt lists, and never more than ten requests in all. They come from our published user agent and a fixed set of static IP addresses listed on our agent page, and robots.txt is honoured for every one. A result is kept for ten minutes, and we cap how often any one store is checked across all our free tools, however many people ask.
See how our own agent identifies itself, and how to opt out, on our agent page.
A tidy file is not the thing that stops an agent
A scan walks your store the way an AI shopper would and tells you where it stalls on the way to checkout. One free scan per store, per email address, every 30 days.