llms.txt validator: check it before you trust it
Paste the file, or point this at a domain and it will read the live one. Most files that fail do not fail on formatting — they fail because nothing is being served at the root at all.
Nothing is uploaded when you paste — the check runs in your browser.
The three ways a published llms.txt is doing nothing
- It is at an address nothing will look at. The conventional place is
example.com/llms.txt. Since v2 a subpath works too —/docs/llms.txtcovers the pages under/docs/— but/pages/llms-txt, which is what several builders produce, covers nothing and is not named llms.txt. - It is served as HTML. Several builders will only give you a web page, wrapped in a full document with your nav and footer around it. What reads the file gets your site chrome and no structure.
- It is stale. It points at pages you moved, or omits the ones you have added since. A file that lists four dead URLs is worse than no file: it is a curated index that curates wrongly.
Then check the thing that matters more
A perfect llms.txt on a site that is blocked from the crawlers is still invisible. Run the AI crawler checker next — it takes one click and answers a bigger question. Or get the full report and find out which buyer questions you are losing, and to whom.
FAQ
Questions
- Why read the live file instead of pasting it?
- Because the most common failure is not a formatting mistake — it is that the file is not being served as a file. Hosted builders will happily give you a page called llms.txt that comes back as HTML wrapped in your nav and footer, which looks right in the editor and is useless to anything reading it. Reading the live URL is the only way to catch that.
- What counts as an error versus a warning?
- An error breaks the structure: no H1, an H1 that is not the first line, a second H1, a section with no name, a summary above the H1, or an item in a section that is not a link or has no title. A warning is something that parses but makes the file less useful — a missing summary, relative URLs, prose inside a section, an empty section.
- My file has no summary. Does that matter?
- It is the single most valuable line in the file. The blockquote under the H1 is what something reads first to decide what this site is; without it the reader has to infer your business from a list of link titles.
- Should the links be absolute URLs?
- Yes. The file gets read away from your site — copied, cached, passed through a pipeline — and a relative path has nothing to resolve against at that point.
- Is there an official specification?
- llmstxt.org, currently at v2 (August 2026). The structure: an H1 with the site or project name — the only required part — then a blockquote summary, then any number of non-heading markdown sections, then H2 sections holding lists of links, each link optionally followed by a colon and a note. This tool checks against that. Two v2 changes worth knowing: the file can sit at a subpath and covers the pages under it, and the Optional section no longer has a mechanical meaning now that the context-expansion tooling that read it is gone.
- Do you store the file?
- No. A pasted file never leaves your browser. A fetched one passes through our server to get around the browser's cross-origin rules and is not written down.
See who the engines recommend instead
The prompts your client loses, who wins them, and the tasks that would change it.