llms.txt

llms.txt is a single Markdown file that hands an AI model a clean, curated map of your site: who you are, and the handful of pages that actually matter. Instead of making the model wander your whole site and guess, you give it the short list.

One caveat up front: llms.txt is a community proposal, not an accepted standard, and no major AI company has confirmed it reads the file in production. It is a low-effort bet, not a guaranteed win.

What it is

llms.txt was proposed by Jeremy Howard of Answer.AI in September 2024. It is a Markdown file you host at the root of your domain (yourdomain.com/llms.txt). The problem it solves: an HTML page is full of navigation, ads, and scripts that waste a model’s limited context. A curated Markdown index gives it the signal without the noise.

The format is deliberately simple. Only the first line is required:

• An H1 with your site or business name. This is the only required part.
• An optional blockquote summary right after it.
• Optional paragraphs with more detail.
• H2 sections, each a Markdown list of links in the form [name](url): short note.
• An optional section literally titled ## Optional, for links a model can skip when it needs to keep things short.

Why AI engines care

In theory, a model that respects llms.txt can read one short file and immediately know your name, what you do, and where your most important pages are, without crawling and parsing dozens of cluttered HTML pages.

Adoption is still thin. Google has said directly that it does not use llms.txt. OpenAI and Anthropic have not confirmed they consume it either, though Anthropic does publish one for its own docs. In server-log analyses, AI crawlers rarely request the file today.

It can still be worth publishing. It is cheap to make, low risk if you keep it accurate, and the ecosystem is young and may adopt it. Writing it also forces you to name your most important pages clearly, which is useful on its own.

How to set it up

Create a file named llms.txt, keep it curated, and host it at your domain root.

# Acme Analytics

> Acme Analytics provides AI visibility consulting and audits
> for B2B software companies.

We help teams get understood, mentioned, and cited by AI engines like ChatGPT,
Claude, Perplexity, and Google's AI answers.

## Core pages

- [What we do](https://acme-analytics.com/services): Our audit and consulting offers
- [Pricing](https://acme-analytics.com/pricing): Plans and what each includes
- [Case studies](https://acme-analytics.com/work): Proof and measured results
- [About](https://acme-analytics.com/about): Who we are and how to reach us

## Optional

- [Blog](https://acme-analytics.com/blog): Background reading, safe to skip

Host it so it is reachable at https://yourdomain.com/llms.txt. There is also a related convention, /llms-full.txt, which holds your full content concatenated into one Markdown document rather than a list of links. It is a de-facto practice, not part of the written spec, so it is optional.

To make the file actually useful instead of a thin link dump, see the llms.txt content quality guide.