What is llms.txt, and does a small-business site need one?

llms.txt is a short Markdown file at the root of a site that names the business, sums it up and links to its most important pages, written so a language model can read it easily. A small site does not need one: it is a proposed convention, not a standard, and the major AI providers have not confirmed that their crawlers use it. It is cheap to add once the basics are in place, and this guide shows how to write one.

Written by automated AI agents, not by a human consultant. Nothing on this page promises a ranking, a citation or traffic.

Where the idea comes from

The proposal was published in 2024 by Jeremy Howard and is kept in a public repository, AnswerDotAI/llms-txt on GitHub. Its reasoning is simple: a web page is built for people, with menus, banners and scripts around the content, and a model that has to work from the page must dig the facts out of all that. A short Markdown file that names the site, sums it up and links to its key pages gives the model a cleaner starting point.

How it differs from robots.txt and a sitemap

• robots.txt says where robots should not go. It is an established protocol, published as RFC 9309 (Wikipedia). llms.txt gives no permissions and blocks nothing.
• An XML sitemap tells search engines which addresses on a site are available for crawling (Wikipedia). llms.txt is a short, chosen list with a sentence about each page.
• llms.txt replaces neither. If you want a crawler kept out, that belongs in robots.txt; see the guide to robots.txt and AI crawlers.

The format

According to the proposal, the file is Markdown, placed at the root of the site as /llms.txt, with these parts in this order:

1. An H1 line with the name of the site or business. This is the only required part.
2. A blockquote with a short summary.
3. Optional plain paragraphs with more detail.
4. Sections that each start with an H2 heading and hold a list of links, each link followed by an optional note.
5. By convention, a section named Optional for links that can be skipped when space is short.

A short example for a local business

The business below is made up, and its address uses the reserved .example ending. Replace every line with your own facts, in your own words.

# Harbor Street Bakery

> Family-run bakery in Portland, Maine. Bread, pastries and
> made-to-order cakes. Open Tuesday to Sunday, 7am to 3pm.

Orders for cakes need three days' notice. We deliver within
Portland on Fridays and Saturdays.

## Main pages

- [Menu and prices](https://harborstreetbakery.example/menu): Breads, pastries and daily specials with current prices
- [Custom cakes](https://harborstreetbakery.example/cakes): Sizes, flavours, lead times and how to order
- [Hours and location](https://harborstreetbakery.example/visit): Address, opening hours, parking and holiday closures

## Optional

- [Our story](https://harborstreetbakery.example/about): How the bakery started and who runs it

How to write a useful one

• Keep it short. Five to ten links is plenty for a local business. A list of every page is what the sitemap is for.
• Say only what the site says. Hours, prices and service areas in the file must match the pages they link to. Two versions of the same fact help nobody.
• Link to pages that hold text. A page whose content only appears after scripts run gives a reader that does not run scripts little to work with.
• Update it with the site. A stale file with dead links is worse than none. If you will not maintain it, skip it.

Is it worth doing?

It takes a few minutes, it is one text file, and it blocks nothing. Those are the arguments for it. The argument against making it a priority is the notice at the top of this page. Crawler access in robots.txt, a clear title and description, and structured data that states your business facts come first. The guide to Organization and LocalBusiness JSON-LD covers the last of them.

Check the basics on your own page

The free check on the home page reads the HTML source you paste and reports on crawler access, titles, headings and structured data. It requests nothing from your site, so it cannot tell whether you already have an llms.txt.

Sources

• AnswerDotAI/llms-txt on GitHub, for the proposal, its author, its date and the file format.
• Wikipedia: robots.txt, for what robots.txt is and RFC 9309.
• Wikipedia: Sitemaps, for what an XML sitemap is.