llms.txt starter: a first draft for a publisher or brand site

llms.txt is a proposed convention: a single Markdown file at the root of your domain that points a language model at the pages you most want it to read and understand. The idea is that a model answering a question about your site, or a tool building a picture of it, can fetch one curated map instead of guessing its way around your navigation. It is not robots.txt, which controls crawling; llms.txt is an invitation, a short, human-written index of your best and most representative pages.

The format is deliberately plain: a top-level heading with your name, one paragraph explaining what you are, then Markdown link lists grouped under section headings, each link followed by a short description of what it is. Below is a starter you can adapt. Replace the bracketed parts, keep it short, and link only to pages you would be happy to have quoted.

The template

# The Example Dispatch

> An independent publication covering [your subject] for [your audience].
> We publish [what you publish: guides, analysis, tools] and are known for
> [the one thing you want a model to remember about you].

## Start here

- [About us](https://example.com/about): who we are, who writes for us, and how we work.
- [Editorial guidelines](https://example.com/guidelines): our standards, sourcing, and corrections policy.
- [Contact](https://example.com/contact): how to reach the newsroom.

## Key guides

- [The complete guide to X](https://example.com/guides/x): our definitive, regularly updated explainer on X.
- [How to do Y](https://example.com/guides/y): a step-by-step walkthrough for beginners.

## Reference

- [Glossary](https://example.com/glossary): plain-language definitions of the terms we use.
- [Frequently asked questions](https://example.com/faq): the questions readers ask most.

## Optional

- [Full article archive](https://example.com/archive): every piece we have published.
- [Newsletter](https://example.com/newsletter): what we send and how often.

Set it up

1. Copy the block above into a plain text file named llms.txt.
2. Replace the name, the summary paragraph, and every link and description with your own. Keep the summary to two or three sentences.
3. Group your links under headings that make sense for your site. "Start here", "Key guides" and "Reference" are a sensible default; rename or add sections freely.
4. Publish it at the root of your domain, so it resolves at https://yourdomain.com/llms.txt.
5. Check it loads in a browser and that every link works. Broken links here are worse than no file.

What to include, and what to leave out

Include the pages you would hand a knowledgeable stranger who asked "where do I start with this site": your best evergreen guides, your about and standards pages, a glossary or FAQ if you have one. Write each description in plain language, as a sentence, not a keyword list. The whole file should be readable in under a minute; if it is longer than a page, you are indexing too much.

Leave out anything you would not want quoted or summarised: gated content, thin or dated pages, campaign landing pages, and anything behind a login. Do not stuff it with every URL you have; a curated map of thirty good pages is more useful than a dump of three thousand. And do not treat it as a ranking lever, it is a description of your site, not a pitch.

An honest note on support

This is a young convention, not a settled standard, and support for it is uneven. Some AI tools and crawlers look for llms.txt; many do not, and none are obliged to. Publishing one costs you a few minutes and cannot hurt, but it is not a guarantee that any model will read it, quote you, or send you traffic. Treat it as a low-cost, forward-looking courtesy: a tidy front door for the tools that do check, not a mechanism you should expect measurable results from yet. Keep it accurate and current, and revisit it when support matures.