THE DEFINITION
What is llms.txt?
llms.txt is a plain-text file, written in Markdown, that sits at the root of a website — yourdomain.com/llms.txt. It gives AI models a short, curated map of what the site is and which pages matter, in a form a model can read directly instead of crawling and parsing everything.
Think of it as the summary you would hand a researcher who has ten seconds: here is what this organisation does, and here are the pages worth reading. It is cheap to write, widely adopted in developer tooling, and — this is the part most explanations skip — entirely unverified.
Published 2026-07-31 · Grounded AEO · groundedaeo.com
→ An H1 with the site or company name.
→ A short blockquote summarising what the organisation actually is, in plain language.
→ Markdown link lists grouping the pages that matter, under headings a model can use as structure.
→ Often a plain “Key facts” block — pricing, platforms, what the product is and is not.
→ Optionally an llms-full.txt alongside it, carrying the expanded content rather than only links.
The two files are constantly confused, and they answer opposite questions. robots.txt is exclusion — it tells crawlers where they may not go. llms.txt is inclusion — it tells a model what is worth reading and how the site hangs together.
One restricts, the other offers. They are complementary, and most sites should have both: robots.txt to allow the AI crawlers you want, llms.txt to tell them what matters once they arrive.
Anyone can write anything in an llms.txt file. It is a self-declaration served from your own server: no proof of domain control, no signature, nothing a model can check. A carefully maintained file and an invented one are indistinguishable to whatever reads them.
That is not a flaw so much as a scope. llms.txt was designed to organise, not to prove. It does its job well. It simply does not answer the question an answer engine has to ask before repeating a claim as fact: how do I know this is true, and who said so?
Worth being equally honest about adoption. Developer tools and agents fetch llms.txt routinely; the major answer engines have not committed to it as a grounding input, and none has published a guarantee that it reads the file. It is a low-cost, well-adopted convention — not a lever with measured output.
A v=AEO1 record is not a replacement for llms.txt, and anyone telling you to swap one for the other has the relationship wrong. They sit at different layers.
→ llms.txt — your own map of your own content. Unsigned, unverified, served by you.
→ v=AEO1 — a signed DNS record pointing at a verified fact catalog. Proof of domain control, and facts an engine can attribute.
Keep the llms.txt. Add the record when you need the claims in it to carry provenance. The open specification lives at aeorecord.org.
Related: What is AEO? · What is GEO? · AI SEO
What is llms.txt?
llms.txt is a plain-text file, written in Markdown, placed at the root of a website — yourdomain.com/llms.txt. It gives large language models a short, curated map of what the site is and which pages matter most, in a form a model can read directly without crawling and parsing the whole site.
How is llms.txt different from robots.txt?
They answer opposite questions. robots.txt is a set of exclusions — it tells crawlers where they may not go. llms.txt is an inclusion: it tells a model what is worth reading and how the site is organised. One restricts access, the other offers a guided summary. They are complementary, and a site can sensibly have both.
What goes in an llms.txt file?
A single H1 with the site or company name, a short blockquote summarising what it is, and then Markdown link lists grouping the most important pages under headings. Many sites add a plain "Key facts" section. The format is deliberately simple: it is Markdown so a model can read it, and short so it fits comfortably in context.
Do AI models actually read llms.txt?
Adoption is real but uneven, and it is worth being precise about this rather than overselling it. Some agents and developer tools fetch it routinely; the major answer engines have not committed to it as a ranking or grounding input, and no engine has published a guarantee that it reads the file. Treat it as a low-cost, well-adopted convention rather than a lever with proven output.
Is llms.txt an official standard?
No. It is a widely adopted convention proposed publicly and picked up by developer tooling — not a standard published through a standards body such as the IETF or W3C. That is not a criticism: robots.txt operated the same way for roughly twenty-five years before it was finally written up as RFC 9309.
Does llms.txt prove anything about my business?
No, and this is the limit worth understanding. Anyone can write anything in an llms.txt file — it is a self-declaration on your own server, with no verification of domain control and no signature. A model reading it has no way to distinguish a carefully maintained file from an invented one. It is useful for organising and pointing; it carries no proof.
How does llms.txt relate to a v=AEO1 record?
They do different jobs and work well together. llms.txt is an unsigned map of your content, written by you and served from your own site. A v=AEO1 record is a signed, DNS-anchored pointer to a verified fact catalog — it exists to answer the trust question llms.txt does not attempt. Keep your llms.txt; the record adds provenance on top of it.
Where should llms.txt be placed?
At the root of the domain, served as plain text at yourdomain.com/llms.txt. Some sites also publish an llms-full.txt containing the expanded content itself rather than just links, for models that will read more.
See what AI engines can actually find about your business
A free audit checks your llms.txt, your structured data, your crawler access and whether a verified record exists — and tells you plainly what is missing.