Browse the docs

What is llms.txt, and do you need one?

What an llms.txt file is, what it does and does not do for AI answers, how to write one, and how to check yours for free.

On this page

An llms.txt file is a short Markdown file at the root of a website, at /llms.txt, that tells AI tools what the site is and which pages are worth reading. It is a proposed convention, not a standard or a crawler rule. It is cheap to write, but robots.txt and readable pages matter more.

What llms.txt is#

The format was proposed at llmstxt.org in 2024. The idea is simple. A website is built for people and browsers, with menus, scripts, and layout. An AI tool reading it wants the opposite, a plain summary and a list of the pages that matter.

The file has a fixed shape.

  1. A heading with the site's name.
  2. A one-paragraph summary, written as a quote.
  3. Sections of links, each a Markdown list of a page and what it covers.

Behavior's own llms.txt, as an example#

Behavior publishes one at behaviorlayer.com/llms.txt. The server builds it from the live plans, public pages, and docs, so it never goes stale. Here is its shape, shortened.

# Behavior

> Behavior is an AI visibility tracker. It shows whether ChatGPT, Gemini,
> Google AI Mode, and AI Overviews recommend a business or its competitors,
> the sources they cite, and what to fix.

## Plans

- Free: $0, with a one-time $1 of checks. ...
- Pro: $119 a month. ...

## Pages

- [Pricing](https://behaviorlayer.com/pricing): every plan and limit
- [Free AEO checker](https://behaviorlayer.com/aeo): whether AI search crawlers can read a website

## Docs

- [Set up AI visibility tracking](https://behaviorlayer.com/docs/visibility-getting-started): ...

Notice what it leaves out, such as sign-in pages, drafts, and anything private. It lists only pages a stranger should read.

What llms.txt does not do#

It does not let crawlers in or keep them out. That is robots.txt's job, and a crawler blocked there will not read your llms.txt either.

It does not guarantee that any AI engine reads it, cites you, or names you. The AI companies decide what their crawlers fetch, and none of that is promised by publishing a file. Treat llms.txt as a helpful summary for tools that look for one, not as a ranking lever.

The evidence agrees. Two large studies published in 2026 found no difference in how often AI answers cite sites with an llms.txt file and sites without one. SE Ranking looked at about 300,000 domains, and Trakkr at 337,000 citations, where sites with the file averaged 6.8 citations and sites without it 6.7. If you want to be named more often, the reviews, comparison pages, and lists that answers cite matter far more.

How to write one#

Keep it short and true.

  • Summary. One paragraph in plain words that says what you sell, who it is for, and where. Use the words your buyers use.
  • Key pages. Products or services, pricing, locations, and the pages that answer the questions buyers ask. A line each on what the page covers.
  • Facts that change. If you list prices or plans, generate the file from the same data as your site, so the two never disagree.
  • Nothing private. No account pages, no staging links, no drafts.

Save it as plain text at https://yoursite.com/llms.txt, and make sure robots.txt does not block that path.

robots.txt matters more#

Whether AI can cite your pages is decided first by robots.txt. A search crawler such as OAI-SearchBot (ChatGPT search), Claude-SearchBot, or PerplexityBot that robots.txt shuts out cannot read the page, so that engine cannot cite it. Training crawlers such as GPTBot and ClaudeBot are a separate choice, and blocking them costs no citations. Let AI crawlers read your site lists every crawler and what to allow.

After robots.txt comes the page itself. Vercel's analysis of AI crawler traffic (December 2024) found that the major AI crawlers, GPTBot and ClaudeBot among them, do not run JavaScript. So a page whose text only appears after scripts run may look empty to them.

Check yours for free#

The free AEO checker reads your robots.txt, looks for llms.txt and a sitemap, and reads up to twenty pages the way AI crawlers do. It needs no account and asks no AI anything.

To see whether AI answers actually name you, run the free check or compare plans to track it every day.

Updated Nov 10, 2026