What Is the llms.txt Standard? The Format Spec, Explained

"Standard" is doing a lot of work in that phrase. llms.txt has no ISO number, no W3C recommendation, and no IETF RFC behind it — it's a community convention, documented at llmstxt.org, that a growing number of sites have chosen to adopt. This page is the deep dive on that spec specifically: what it requires, what it leaves open, and where it's actually written down. For the broader picture — why the file exists and adoption data — start with What Is llms.txt?. If you want real files instead of the spec, see llms.txt examples.

Who actually maintains the "standard"

No governing body owns llms.txt the way the W3C owns HTML, or the way the Sitemaps protocol has a formal XML schema at sitemaps.org. The specification lives on llmstxt.org (accessed 2026-09-01), a community-run site that documents the format and links to independent directories of sites that have published one. Nothing requires a site to follow it, and nothing requires an AI tool to read it. That voluntary-on-both-sides structure is exactly why calling it a "standard" is a little generous — it's closer to a widely copied convention, the same category robots.txt started in before search engines converged on honoring it.

The exact format, piece by piece

Per llmstxt.org's own specification, an llms.txt file is built from these elements, in this order:

• An optional byte-order mark (BOM) — a low-level encoding marker most people never think about and don't need to add manually.
• An H1 heading with the project or site name — the only element the spec actually requires. Nothing else is mandatory.
• A blockquote with a short summary — optional per the letter of the spec, but close to universal in practice, since it's the fastest way to give a reader, or a model, the gist.
• Zero or more free-form Markdown sections — paragraphs of additional detail, with no required heading names or content.
• Zero or more H2-delimited sections containing file lists — each a Markdown list of links, formatted as [name](url) followed optionally by a colon and a short note.

The spec's own template, reproduced here, is exactly this shape:

# Title

> Optional description goes here

Optional details go here

## Section name

- [Link title](https://link_url): Optional link details

Notice how much of that is "optional." The one line that isn't is the H1.

What the spec deliberately does not define

This is the part most guides skip, and it's the actual answer to "is my file following the standard correctly":

• No length limit. Nothing caps how many links or sections a file can have.
• No required section names. "Docs," "Guides," "Product" — none of these are reserved words. Name sections whatever makes sense for your content.
• No file-size cap or hosting requirement beyond the root path. The convention is to serve it at /llms.txt as plain text, but the spec text itself doesn't mandate a MIME type or maximum size.
• No validation schema. Unlike sitemap.xml, which search engines can mechanically validate against an XML schema, there's no equivalent schema for llms.txt. A file either has a parseable H1 and links, or it doesn't — there's no formal pass or fail beyond that.
• No required llms-full.txt companion. The expanded, full-text companion file some sites publish is a common practice, not a spec requirement.

Because there's no formal validation schema, "is there an official llms.txt validator" doesn't have a clean yes. There's no single tool a governing body maintains or endorses. Some third-party checkers exist independently of the spec itself, and GeoReady's own audit checks structural validity — H1, blockquote, sections, links — as one signal inside its broader llms.txt category, alongside checking that the file is actually reachable and not blocked. Neither is "the" official validator, because the spec never defined one to certify against.

Where the standard is actually documented

llmstxt.org (accessed 2026-09-01) is the specification's home. From there, it points to three independent directories that list sites known to have published an llms.txt file: llmstxt.site, directory.llmstxt.cloud, and llmstxthub.com. None of these directories are official registries in the sense of, say, a domain registrar — they're community-maintained lists, useful for browsing real examples but not authoritative about who "correctly" implements the format, since there's nothing to certify against.

How this compares to the other root-level files you already have

• robots.txt has decades of de facto standardization behind it — no formal spec body either, but near-universal crawler compliance through convention. The closest historical parallel to where llms.txt is now.
• sitemap.xml does have a semi-formal spec at sitemaps.org, with an actual XML schema search engines validate against and a submission mechanism through Search Console.
• llms.txt sits earlier in that maturity curve than either — younger, less universally read, and with no submission or validation mechanism at all. Complementary to both, per What Is llms.txt?, not a replacement for either.

Does following the spec exactly matter?

Practically, yes for the required part and no for the rest. Tools that parse llms.txt — including GeoReady's own llms.txt generator — expect the H1-then-links shape, so straying from it risks a file that reads as malformed to anything trying to parse it programmatically. But the real llms.txt examples fetched for that companion page show plenty of variation beyond the required minimum — extra sections, agent-usage notes, companion files — none of which breaks the format, because the spec never forbids additional content. Follow the required skeleton exactly. Treat everything else as a judgment call about your own site.

Common misconceptions about the "standard"

• "It's enforced somehow." It isn't. No crawler is required to read it, and no penalty exists for skipping it or getting it wrong.
• "It's equivalent to robots.txt." It isn't — robots.txt controls access; llms.txt orients readers to content. Neither substitutes for the other.
• "There's an official certification or validator." There isn't one maintained by a governing body — see above.
• "Google uses it for ranking." It doesn't. No search engine has confirmed llms.txt as a ranking input; treat any claim otherwise as unverified until a vendor says so.