What Is llms.txt? Format, Example, and When to Use It

llms.txt is a proposed Markdown guide to a website's information for AI agents. It provides context and links to useful resources. Publishing one is optional: it does not replace robots.txt or a sitemap, and Google does not require a special AI text file for AI Overviews or AI Mode.

An agent helping someone use your product needs to find the right instructions. Your website might spread those instructions across setup pages, policy documents, and old release notes. A carefully maintained starting point can make that task easier to organize. That is the practical reason to consider llms.txt before discussing rankings.

The useful question is whether you have information to organize and a reader or tool that can use it. Adding another file without either creates another maintenance obligation.

What does an llms.txt file contain?

The llms.txt proposal describes a Markdown file with a project or site name as its H1. That heading is the only required section. The suggested order then includes a short blockquote summary, explanatory text, and H2 sections containing lists of links with optional descriptions.

Use the explanations to distinguish documents that sound similar. A developer needs to know whether a link explains installation, current API behavior, or historical changes. A customer-service agent might need eligibility rules rather than marketing copy. Choose the resources around those tasks.

This is an information guide. Its contents should remain consistent with the pages people can read and the policies your organization actually follows.

A small llms.txt example

Here is an original illustration for a fictional appointment tool. The example.com addresses are placeholders, not working documentation:

# Harbor Desk

> Public documentation for a fictional appointment tool.

Current instructions describe the supported service. Older releases are historical.

## Product

- [Setup](https://example.com/docs/setup.md): Connect an appointment calendar.
- [Limits](https://example.com/docs/limits.md): Plan restrictions and supported regions.

## Optional

- [Release notes](https://example.com/docs/releases.md): Changes organized by date.

Replace the name, summary, and URLs with your own approved material. Avoid a long list of every landing page. A short explanation of which document answers which question is more useful than repeated claims that your product is the best.

The proposal recommends links to content agents can readily consume, such as clean Markdown versions. Do not invent .md URLs by adding an extension to pages that have no Markdown endpoint.

Where does the file go? What changed in v2?

The proposal's August 2026 revision clarifies placement: /llms.txt can cover a site, while /docs/llms.txt covers that path. Where scopes overlap, the proposal calls for the most specific file.

It also describes discoverability links. For example, a documentation page could include:

<link rel="describedby" href="/docs/llms.txt">
<link rel="alternate" type="text/markdown" href="/docs/setup.md">

The first points to the covering guide; the second identifies a Markdown alternative. HTTP Link headers can also provide these relationships. They are proposal conventions, so check how your intended consuming tool handles them. A published link is not proof that every assistant follows it.

An Optional section remains a convention for secondary resources. Do not assume its heading forces a tool to exclude those links.

llms.txt, robots.txt, and sitemaps do different jobs

Google's robots.txt documentation describes crawler access rules. Those rules are not a secure way to protect private information, and disallowing a page does not necessarily remove its URL from Search. Keep access and indexing decisions in the appropriate controls.

A sitemap provides search engines with information about site URLs and their relationships. It helps discovery without guaranteeing that every listed page will be crawled or indexed. Maintain it alongside ordinary navigation.

An llms.txt file supplies a curated reading route. It should not be treated as authorization to access private pages or as a command that overrides crawler policies. For a separate explanation of crawler choices, see our guide to GPTBot and OAI-SearchBot.

Does llms.txt improve SEO or AI citations?

Google's AI-feature guidance says new AI text files are not required for AI Overviews or AI Mode. Supporting pages need to be indexed and eligible for a Search snippet. Meeting the requirements still does not guarantee selection.

Do not turn evidence that a tool reads your file into a claim that search engines reward it. Those are different observations. If important pages have access problems or unsupported business claims, address those first using the AI search audit checklist.

For a documentation team, a reasonable experiment is narrower: can a selected tool find the current setup guide from this starting point and answer a defined question accurately? Record the tool, question, version of the file, resources retrieved, and mistakes. That tests a concrete use case without promising broader discovery.

How to publish and maintain a useful version

Start with a small document set and a named owner. Review the summary against the linked sources, remove superseded material, and check that public URLs return the intended content rather than a sign-in page or an application shell.

Open the file directly after deployment. Follow each link, including Markdown alternatives, and verify that important limits and instructions survive the conversion. Use fictional or approved public examples; exclude customer records, credentials, and internal procedures.

Review the file when documentation moves or policies change. If you measure broader results, keep retrieval observations separate from citations, referral visits, and qualified inquiries. Our AI referral measurement guide explains that distinction.

For a small business with a few clear service pages, maintaining those pages may deserve attention first. Add llms.txt when the information guide has a defined purpose, an owner, and a manageable upkeep cost.

FAQ

Is llms.txt mandatory for Google AI Overviews?

No. Google says special AI text files are not required for AI Overviews or AI Mode. Focus first on useful pages, access, indexing, and snippet eligibility. An llms.txt file is an optional publishing choice, not an admission requirement.

Must llms.txt be at the domain root?

No. The current proposal also permits files under paths such as /docs/llms.txt. Choose placement that matches the material you control, and check whether your intended tool understands that scope. Do not assume all clients implement the proposal identically.

Do I need Markdown versions of every page?

No. The proposal recommends clean Markdown resources for agents, but an initial file can focus on selected public documents. Only link Markdown endpoints that actually exist, and verify their accuracy against the public originals rather than generating broken addresses.

Can llms.txt block training or protect private content?

Do not use it for either purpose. It is a public information guide, not an access-control mechanism. Review the relevant provider's crawler controls separately, and protect confidential material through authentication and appropriate server permissions.

How can I tell whether the file is useful?

Define a task before testing: for example, finding the current setup instructions and answering one supported configuration question. Record which sources the selected tool retrieved and whether its answer was correct. A successful task is useful evidence for that workflow, not proof of increased rankings or citations.