Plenty of tools generate an llms.txt — almost none tell you whether yours is right. Enter your website and this checker fetches your live /llms.txt, validates it against the llmstxt.org spec line by line, and scores it 0–100 with a finding-by-finding breakdown, plus whether llms-full.txt exists. Free, no signup.
The checker takes any public website URL and requests /llms.txt and /llms-full.txt at the root of your domain — the exact locations the llmstxt.org proposal defines.
Both files are fetched the way a crawler would see them, including the most common silent failure: a server that returns an HTML page with a 200 status instead of a real plain-text file (a soft 404). That counts as missing, because it is.
The file is parsed and validated against the llmstxt.org format: an H1 title as the first line, a blockquote summary, H2 sections containing link lists, link URL quality, malformed list lines, file size, and stray HTML markup.
You get a 0–100 score and a pass/warn/fail finding for each check with the exact lines that need fixing, plus section, link, and size stats — so a valid file is a two-minute edit away, not a guessing game.
One number for how closely your file follows the llmstxt.org spec, weighted toward the rules that matter most: the H1 title and the linked H2 sections that make the file useful to an AI assistant.
Every check reported individually with a plain-English message — including examples of malformed link lines and links assistants may not resolve — so you know exactly what to change.
Section count, link count, and file size, with a warning when the file grows past the ~50KB a curated index needs — long content belongs in llms-full.txt.
Whether the optional companion file exists and how large it is. It isn’t scored — it’s the spec’s home for full page content once your curated llms.txt is in place.
llms.txt is a proposed standard (llmstxt.org) for a plain markdown file at /llms.txt that gives AI assistants a curated map of your site: an H1 title, a one-line blockquote summary, and H2 sections of links to your most useful pages. Its companion, llms-full.txt, can carry full page content for assistants that want everything.
It starts with "# Your Site Name", then a one-line summary as a blockquote ("> What you do"), then H2 sections like "## Docs" containing link lists in the form "- [Page title](https://yoursite.com/page): short description". An "## Optional" section is the spec-blessed place for links assistants can skip when context is tight.
Each spec rule carries a weight, and the weights sum to 100. The heaviest are the H1 title and having at least one H2 section with a valid markdown link list; the summary blockquote, link URL quality, list syntax, file size, and absence of HTML markup make up the rest. A passing check earns its full weight, a warning earns half, a failure earns nothing.
The most common cause is a soft 404: your server (often a single-page app with a catch-all route) returns your normal HTML page with a 200 status for /llms.txt instead of the actual text file. Assistants and crawlers see the same thing this checker does, so the fix is making /llms.txt serve the real plain-text file.
Honestly: adoption by AI providers is still emerging, and no major assistant has committed to reading llms.txt the way search engines committed to robots.txt or sitemaps. Treat it as a forward-looking, low-cost signal — a clean map of your best content that costs one file to publish — not a guaranteed ranking factor. What you can measure today is whether AI assistants mention your brand, which is what Magister’s visibility tracking does.
llms.txt is the curated index — short, links-first, meant to fit comfortably in a model’s context. llms-full.txt is the expanded companion that can inline full page content for assistants that want everything in one fetch. That’s why this checker warns when llms.txt itself balloons past ~50KB: long content belongs in the full file, not the index.
This check is one signal. The free audit grades your whole site for AI search and shows the plan Magister would run to fix what it finds.