llms.txt Examples & Templates: What to Put in Yours (and How to Validate It)

llms.txt is a machine-readable table of contents at /llms.txt: a short markdown summary of who you are plus the handful of pages an AI system should read first. It does not grant crawl access (that is robots.txt) and it is not a ranking signal; it is a routing file that makes retrieval cheaper and more accurate. After running one on this site for over a year (see our 90-day retrospective), here are the templates that earn their keep and the validation checks that catch broken ones.

The anatomy that works

# Product Name
> One-paragraph summary: what it is, for whom, what it does, pricing model.

## Core pages
- [Pricing](https://example.com/pricing): plans, limits, what each tier includes
- [Product overview](https://example.com/product): features, platforms, integrations
- [FAQ](https://example.com/faq): billing, data, cancellation

## Docs
- [Getting started](https://docs.example.com/start): install, first call, auth
- [API reference](https://docs.example.com/api): endpoints and schemas

# Product Name
> One-paragraph summary: what it is, for whom, what it does, pricing model.

## Core pages
- [Pricing](https://example.com/pricing): plans, limits, what each tier includes
- [Product overview](https://example.com/product): features, platforms, integrations
- [FAQ](https://example.com/faq): billing, data, cancellation

## Docs
- [Getting started](https://docs.example.com/start): install, first call, auth
- [API reference](https://docs.example.com/api): endpoints and schemas

Three rules behind the template:

1. The blockquote is the money line. It is the text most likely to be lifted verbatim when a model describes you. Write it like a directory listing: product, audience, function, price model — one paragraph, no slogans.
2. Five to fifteen links, not fifty. The file competes for attention in a retrieval step; a wall of links dilutes it. Choose the pages that answer buyer questions, not your sitemap.
3. Every URL absolute and live. Relative paths and 404s are the two most common ways an otherwise good file becomes dead weight.

Variants by site type

• Product site: overview, pricing, comparison/vs pages, changelog, FAQ.
• Docs site: getting started, auth, the three most-used endpoint groups, migration guides.
• Blog/publisher: the four pillar posts, the about page, and the data reports — skip the individual articles.

What it cannot do

It will not get a blocked crawler in (robots.txt wins), will not create citations where the content is thin, and is ignored by systems that never fetch it. Its value is precision: fewer wrong summaries, fewer "the model described our old pricing" incidents.

Validation checklist

• Returns 200 with text/plain or markdown at https://<domain>/llms.txt (not a 404, not HTML).
• First line is a # Title; a > summary block follows.
• All links absolute, HTTPS, and resolving (script a HEAD request per line in CI).
• Summary matches the canonical description used in directory listings — one fact sheet, everywhere (see how to rank in ChatGPT, lever 2).
• Under ~2 KB. If you need more, you need fewer links, not a bigger file.
• Updated when pricing or positioning changes; a stale llms.txt teaches models your old facts.

FAQ

Does every site need one? Any site that wants to be described accurately by AI systems, yes — it is five minutes of work with the templates above. It matters most for products with complex pricing or frequent renames.

llms.txt or robots.txt first? robots.txt first: access before routing. Our AI crawler guide covers the access rules and the agents behind them.

Should I list competitors' comparison pages? No. llms.txt routes readers to your canonical facts. Comparison pages about you on other domains belong in your mention-earning work, not your routing file.

Running a listing footprint with the same canonical description your llms.txt uses is what makes the facts stick: submit your product on aat.ee, permanent and dofollow across the network.