What is llms.txt, and should you add one?

Short answer: llms.txt is a plain Markdown file you put at the root of your site (yoursite.com/llms.txt) that lists your most important pages with a one-line description of each, so an AI tool can find the right page without crawling everything. It was proposed in September 2024 and is still a proposal: as of November 2026 none of the major assistants has published that it reads the file. It costs about 20 minutes, does no harm, and can help tools that do fetch it. It will not, on its own, get your business named in an AI answer.

What the file is

The proposal came from Jeremy Howard of Answer.AI in September 2024. The idea borrows from robots.txt and sitemaps: one predictable URL where a machine can learn what a site is about and where its key pages live. Unlike a sitemap, it is written for a language model to read, so it is Markdown rather than XML and it carries short descriptions, not just URLs.

The spec asks for four things, in this order:

1. One H1 with the name of the site or project. This is the only required line.
2. A blockquote with a one or two sentence summary.
3. Optional paragraphs with context a model should know.
4. One or more H2 sections, each holding a list of links in the form [Title](URL): description.

A section named Optional is reserved: it marks links a model can skip if it has a tight context budget.

Many sites also publish /llms-full.txt, which inlines the full text of the pages instead of linking to them. That is useful for documentation sites with thousands of pages. For a business site with 20 to 50 pages it is rarely worth maintaining.

Who reads it today

This is the honest part. The file is widely adopted by developer documentation sites and by AI coding tools that fetch docs on request. The consumer assistants are a different story:

• OpenAI, Anthropic, Perplexity and Google have not published support for llms.txt in ChatGPT, Claude, Perplexity or Gemini as of November 2026.
• Google's John Mueller said in 2025 that no AI system he knew of used the file, and compared it to the keywords meta tag.
• Server logs from sites that publish the file show occasional fetches by assorted bots, which is not the same as the file shaping an answer.

So treat it as cheap insurance, not a lever. The assistants still find your business the way they always have: by retrieving pages from a web search, reading them, and checking that the facts agree across your site, your listings and third-party pages. That is the work covered in our AI search optimization guide.

Why add one anyway

Three reasons, in order of weight:

1. It is 20 minutes of work with no downside. A static text file cannot hurt your rankings or your crawl budget.
2. It documents your own priorities. Writing one line per key page forces you to decide which 15 pages matter. Most businesses find thin or duplicate pages in the process.
3. Some tools do fetch it. Retrieval tools built on frameworks that support the convention, and any assistant that adds support later, will get a clean map instead of guessing from navigation menus.

A copy-ready example

Here is a template for a local service business. Replace the names and keep every description to one sentence.

# Northside Plumbing

> Licensed plumbing company serving Austin, TX since 2009. 24/7 emergency service, water heaters, repiping and drain cleaning.

Family owned. Licensed and insured in Texas (license M-12345). Service area: Austin, Round Rock, Cedar Park and Pflugerville.

## Services

- [Emergency plumbing](https://example.com/emergency/): same-day response, 24 hours a day, with flat-rate pricing published up front.
- [Water heater repair and replacement](https://example.com/water-heaters/): tank and tankless, with brands we install and typical costs.
- [Drain cleaning](https://example.com/drains/): hydro jetting and camera inspection, and when each is the right call.

## About

- [About Northside Plumbing](https://example.com/about/): the team, licenses, insurance and service area.
- [Reviews](https://example.com/reviews/): verified customer reviews with dates.
- [Contact](https://example.com/contact/): phone, hours and address.

## Optional

- [Blog](https://example.com/blog/): how-to articles on common household plumbing problems.

Save it as llms.txt at the root of your site, make sure it is served as plain text, and link nothing to it. You can see ours at frontpagely.com/llms.txt.

Do you need a generator?

No. Several "llms.txt generator" tools exist, and most crawl your site and dump every URL into the file. That defeats the purpose. The value is in the curation: the 10 to 30 pages you would hand a new employee on day one, each with a sentence that says what the page answers. Write it by hand, or start from a generator's output and cut it down.

Where it fits in the bigger picture

If you are trying to get your business named when customers ask an AI assistant, llms.txt is near the bottom of the list. The inputs that measurably matter are the ones the assistants retrieve and corroborate: a clear schema markup describing who you are, consistent listings, pages that answer the question a buyer actually asks, and third-party pages that back you up. The free audit shows which of those you are missing and which engines currently name you, so you can spend your 20 minutes on llms.txt and the rest of the afternoon on the things that move the number.

For the other terms in this space, see the AI visibility glossary.