How the checks work

Spec validity answers one question: is this a valid llms.txt? Five checks (E1, E2, E3, E4, E5): the file exists at its path, is served as text, has an H1, its file-list lines are well formed, and it is not an HTML page. A leading byte-order mark is fine. Fail any of them and the verdict is “invalid”.

Quality is Studio's set of recommendations (W1, W2, W3, W5, W6), graded 0–100: start at 100, subtract per hit up to a cap, letter by threshold (A ≥ 90 · B ≥ 75 · C ≥ 60 · D ≥ 40). None of these is required by the spec, and the report says so on every rule.

Information (I1, I2, I3, I4, I5, I6) records facts: robots.txt blocks, llms-full.txt presence, file age, whether pages declare rel="describedby", how many links resolve to markdown, and external links. They change neither verdict.

When nothing readable is served at all (E1 or E5) quality cannot be assessed, so the grade reads F with that reason.

v2 lets an llms.txt live at a subpath and cover the pages under it: /docs/llms.txt describes everything in /docs/, and the most specific file applies. The validator accepts a full URL such as example.com/docs/llms.txt and reports which prefix the file covers. The generator takes an optional path scope such as /docs: it only reads pages under that path and tells you to publish the result at /docs/llms.txt.

v2 also recommends that pages carry a markdown version at the same URL with .md appended or the extension replaced, announced with rel="alternate" type="text/markdown". When a page declares one, or a probed .md sibling answers as text/markdown, the generator links to that instead of the HTML page.

What page do I land on after paying? Straight onto the monitoring page for the domain you were setting up, with a “Subscription active” banner. If you haven't proven you control the domain yet, the verification card (a DNS TXT record or a meta tag, plus a “Check now” button) is the first thing on that page — monitoring starts the moment verification passes. If Stripe sends you back a second or two before we've received its confirmation, the page shows “Confirming your payment…” and updates itself; you don't need to refresh.

How do I get my monitoring updates? Three ways. First, an alert email whenever a weekly check finds something worth knowing: the file is missing, the grade dropped, a link died, your site drifted away from the file, the file itself changed (the email shows the lines that differ), or a documented AI crawler fetched your canary link. You get at most one alert email a day; several findings are bundled into it. Second, a monthly report that lists what we checked even if nothing changed, so you can see the checks are running. Third, your dashboard, where every check, diff and crawler fetch is kept and where the “Regenerate now” button lives.

Which email address do alerts go to? The email on your llms.txt Studio account — the one you signed up with, or your Google account if you used “Continue with Google”. Stripe receipts go to the same address; the email field on the payment page is fixed to it. Change it any time under Account.

• Free tools: 1 generate and 1 validate per day per network address, resetting at 00:00 UTC. No account.
• Generator: robots.txt honoured, up to 60 pages, 2 MB per response, 8 s timeout, 2 requests per second per host, 3 minutes per job.
• Validator: fetches the file, robots.txt and the covering page; probes a sample of 20 links; the file must be under 1 MB.
• Monitor: weekly checks per paid, verified domain; at most one alert email per domain per day; regeneration without the daily limit.

Platform specifics are in the guides (Next.js, WordPress, Webflow, Hugo, Docusaurus), the format itself in the knowledge base, and our own fetcher on /bot.