How to Test and Validate Your llms.txt File Before Publishing (2026 Checklist)

You can test and validate your llms.txt file in about fifteen minutes with no special software — but not the way you might expect. There is no official llms.txt validator, no ping endpoint, and no search console that checks it. Validation is a manual checklist against the specification: correct Markdown structure, exactly one H1, a summary blockquote, working absolute links, and honest expectations about what the file can and cannot do. This guide gives you that checklist, the step-by-step test, and the common mistakes that make an llms.txt file useless.

What the Spec Actually Requires

The llms.txt proposal was published by Jeremy Howard of Answer.AI in September 2024 at llmstxt.org, where it is still maintained as a community proposal — not a ratified web standard. Version 2 of the proposal was published in August 2026. The format is deliberately simple Markdown:

• An H1 with the site or project name. Written as # Your Site Name. This is the only strictly required part.
• A blockquote summary. One line starting with > that states what the site is in a sentence or two. Optional in the letter of the spec, essential in practice — it is the first thing a model reads.
• Optional detail paragraphs or lists. Extra context about the site. These must not use headings — the spec reserves headings for structure.
• H2 sections of links. Groups like ## Documentation or ## Key Pages, each containing list items in the format - [Page Title](https://yoursite.com/page): one-line description of what the reader will find.
• An "Optional" section for secondary links. In version 1, tools were allowed to skip this section to save space; version 2 dropped that rule, so it is now simply a clearly-labeled home for lower-priority pages.

There is also a companion file, /llms-full.txt, which holds the full text of your key pages concatenated into one Markdown document — the whole site in a single fetch. And one thing to get straight before you validate anything: llms.txt is not a crawl-control file. Unlike robots.txt, it allows and blocks nothing, and Google has documented that Google Search does not use it — it neither helps nor hurts visibility in search, AI Overviews, or AI Mode. Its value is with AI assistants and agents that read it directly.

The 12-Point Validation Checklist

Run through these in order. A file that passes all twelve is spec-compliant and genuinely useful; most files fail at least three.

1. The file is served at the domain root. It must live at yoursite.com/llms.txt — not in a subfolder, not at /docs/llms.txt. The location is the whole convention.

2. It returns HTTP 200 as plain text. Fetch the URL. You should get the file itself with a text/plain or text/markdown content type. A 404, a redirect chain, or an HTML error page means the file effectively does not exist for automated readers.

3. Exactly one H1. One line starting with a single #, naming the site or project. Two H1s, or none, both break the expected structure.

4. The blockquote follows the H1. The summary line should come right after the title, before any sections. If your summary is buried three sections down, most readers will never see it.

5. No headings inside detail paragraphs. The spec reserves ## for link sections. If you have explanatory text, keep it as plain paragraphs or bullet lists — no H3s, no H4s.

6. Every link uses an absolute URL. - [Pricing](https://yoursite.com/pricing), never - [Pricing](/pricing). A relative link is meaningless to a model that fetched the file out of context. This is the single most common mistake.

7. Every link works. Click through all of them, or fetch them and check the status codes. Dead links, redirect chains, and pages behind logins make the file worse than useless — they teach readers not to trust it.

8. Every link has a description. The format is [Title](URL): description. The description after the colon is what tells a model why this page matters. A bare link list is just a sitemap with worse formatting.

9. Sections are H2, and sections are curated. Group links under ## headings with clear names. And curate: the file should list your 10–20 most important pages, not all 400. If everything is important, nothing is.

10. The Optional section exists and is honest. Secondary links — changelog, old blog posts, press kit — go under ## Optional. Don't hide your best pages here and don't promote weak ones into the main sections.

11. The Markdown actually parses. Paste the file into any Markdown previewer. If the preview shows broken formatting — a link that doesn't render, a heading that shows raw # characters — fix it. Common culprits: unescaped parentheses inside link URLs, and descriptions containing unescaped square brackets.

12. The content is current. An llms.txt pointing at last year's pricing page or a discontinued product actively misleads. Re-validate on the same schedule you update your sitemap.

How to Test It Step by Step

Here is the full test routine, start to finish:

1. Fetch it as a machine would. Open yoursite.com/llms.txt in a browser and view the page source. What you see in view-source is close to what an automated reader gets — no styling, no scripts, just the text. If you see HTML wrappers, your server is serving the wrong content type or an error page.
2. Check the headers. In your browser's developer tools, look at the response for the llms.txt request. You want status 200 and a plain-text content type. A 301 redirect to add www or https is acceptable as long as it resolves in one hop — chains of redirects are not.
3. Walk the checklist above against the source you just fetched. Count the H1s. Verify the blockquote position. Scan every link.
4. Click every link. All of them. Confirm each lands on the right page, loads without a login, and matches its description. Fix or remove anything stale.
5. Paste it into a Markdown previewer to catch rendering bugs.
6. Read it as a stranger. Hand the raw text to a colleague — or paste it into an AI assistant — and ask: "what does this site do, and which three pages should I read first?" If the answer is wrong or vague, your curation or descriptions need work.
7. Check the companion file if you publish one. Fetch yoursite.com/llms-full.txt and confirm it exists, parses, and doesn't contain content you wouldn't want quoted — because that is literally what it is for.

If building the file by hand feels fiddly, our free llms.txt generator creates a spec-formatted file from your site's pages — then run this checklist against its output before you publish. For the full background on the format, see our complete llms.txt guide, and if you want a ready-made starting point, the free llms.txt template walkthrough covers the no-signup route.

Common Mistakes That Break llms.txt Files

• Relative links. Worth repeating because it is that common: /pricing means nothing outside your site's context. Always absolute.
• Dumping the whole sitemap in. A 300-link llms.txt defeats the purpose. The file is a curated briefing, not an index. Your sitemap already does indexing.
• Treating it as a ranking lever. No major AI engine has publicly committed to reading llms.txt, and Google says its search ignores it. One independent analysis of its own site found four AI engines answering an llms.txt question without citing the site once, and a study of nearly 300,000 domains found no citation effect. Publish it as a cheap, forward-looking extra — ten minutes of work — not as an SEO tactic.
• Marketing fluff in descriptions. "Revolutionary synergy platform" tells a model nothing. "Project management tool with Gantt charts, $12/user/month, free 14-day trial" tells it everything.
• Forgetting it exists. Sites change; the file doesn't update itself. A quarterly review takes five minutes.
• Blocking it in robots.txt. Yes, people do this — usually by accident, with an over-broad disallow. The file that briefs AI readers should itself be readable.

What Validation Can't Do

Be honest with yourself about the limits. Passing this checklist means your file is well-formed and useful to anything that reads it. It does not mean ChatGPT, Claude, or Perplexity will read it, cite you more, or rank you higher — there is no public evidence of any of that today. What a valid file does guarantee: if an agent or assistant fetches it — a coding assistant pointed at your docs, a research agent doing due diligence — it gets a clean, accurate briefing instead of guessing from your HTML. That is a real, if modest, win, and it costs almost nothing.

FAQ

Is there an official llms.txt validator?

No. Unlike XML sitemaps, which have validators and a ping protocol, llms.txt has no official validation service, no schema to check against, and no endpoint to notify. The checklist in this guide — structure, links, freshness — is the validation process. Third-party generators sometimes include basic checks, but nothing authoritative exists.

Does Google read llms.txt?

No. Google has documented that Google Search does not use llms.txt, and that the file neither helps nor hurts visibility or rankings, including in AI Overviews and AI Mode. Don't publish it for Google; publish it for the AI assistants and agents that read websites directly.

What's the difference between llms.txt and llms-full.txt?

llms.txt is the curated map: your site's name, a summary, and links to key pages with descriptions. llms-full.txt is the full meal: the actual text content of those pages concatenated into a single Markdown document, so a model can ingest your knowledge base in one request. Publish the map always; publish the full file when your content is the product — documentation, knowledge bases, reference sites.

How big should the file be?

The spec sets no hard size limit, but the entire point is brevity — a model reading it has a limited context window. Keep it to your most important pages with one-line descriptions each. If your llms.txt is approaching the length of an actual article, you have stopped curating. Ten to twenty links is the healthy range for most sites.

Should I submit llms.txt to search engines or put it in my sitemap?

No. Search engines don't consume it, so there's nothing to submit and no ping endpoint to call. Just serve it at the domain root and keep it fresh. You can mention it in your site's documentation or developer pages so human visitors — the ones pointing AI tools at your site — know it exists.

Can llms.txt hurt my SEO?

There is no documented mechanism by which it could. It's a static text file that search crawlers ignore. The only real risk is indirect: a stale or misleading file briefing an AI agent badly, which is a maintenance problem, not an SEO problem.

Bottom line: There is no official llms.txt validator — validation means checking the file against the spec yourself: root location, HTTP 200, exactly one H1, summary blockquote, absolute working links with real descriptions, curated H2 sections, and fresh content. It takes fifteen minutes, it guarantees nothing about rankings or citations, and it is still worth doing, because a clean briefing beats a guessed one every time an AI reader actually shows up.