llms.txt Guide: Format, Example, and GEO Best Practices

What is llms.txt?

llms.txt is an open proposal first published in 2024 for providing language models with a concise site summary and curated links at inference time. The suggested location is the root path /llms.txt, and the content is Markdown so it can be read by both people and automated tools.

The proposal addresses a practical problem: websites can be large and navigation-heavy. A supporting tool could use a curated file to identify authoritative documentation, product information, policies, examples, or references. That benefit depends entirely on the tool choosing to fetch and interpret the file.

The proposed llms.txt format

The proposal defines a small ordered structure:

1. An optional byte-order mark.
2. An H1 containing the project or site name. This is the only required section.
3. An optional blockquote with a concise summary.
4. Optional Markdown content without headings for important interpretation notes.
5. Zero or more H2 sections containing Markdown link lists.

Each resource entry uses a Markdown link and may include a short description after a colon. A section named “Optional” has a special purpose: its links can be skipped when a smaller context is needed.

A practical llms.txt example

If a tool you need explicitly supports the proposal, the example below shows a concise file for a hypothetical B2B software company. Replace every statement and URL with verified information from your own site.

# Example Analytics

> Example Analytics is a B2B reporting platform for product teams.

Use the product documentation for current capabilities and the
security center for compliance information. Pricing can change;
use the pricing page as the canonical source.

## Product

- [Product overview](https://example.com/product): Core capabilities and use cases
- [Pricing](https://example.com/pricing): Current plans and billing terms
- [Documentation](https://example.com/docs): Setup, API, and integration guides

## Trust and support

- [Security](https://example.com/security): Security controls and compliance
- [Help center](https://example.com/help): Support articles and troubleshooting

## Optional

- [Company blog](https://example.com/blog): Educational articles and updates

Keep the file selective. A sitemap may contain thousands of URLs; the proposal intends llms.txt to point a compatible tool toward a small set of authoritative resources. Do not assume unsupported systems will read or honor it.

llms.txt vs robots.txt vs sitemap.xml

These files are not interchangeable. llms.txt does not grant or deny crawler access. A sitemap remains the scalable discovery mechanism for canonical pages. Neither file replaces helpful page content or accurate structured data.

What should you include?

Choose resources that answer stable, high-value questions about the organization:

• Canonical product or service overview
• Authoritative documentation and API reference
• Pricing, plans, or purchasing guidance when public
• Security, privacy, trust, and compliance information
• Important policies and customer support resources
• Research, datasets, or technical references you maintain

Do not add private URLs, credentials, internal notes, temporary campaign pages, or claims that are not publicly supported. The file is public and should be reviewed like any other published content.

Implementation and QA checklist

1. Define the audience. Decide whether the file primarily supports product research, documentation use, customer support, or another clear task.
2. Select canonical sources. Link only to the pages your team considers authoritative and current.
3. Write a factual summary. State what the organization or project is, who it serves, and what the linked resources cover.
4. Use absolute HTTPS URLs. Make links portable and easy to resolve outside your site.
5. Return plain text or Markdown cleanly. The file should load without authentication, redirects, or an HTML error page.
6. Test links and content. Check response status, destination, duplicates, and whether descriptions still match.
7. Assign an owner. Review the file when products, documentation, policies, or canonical URLs change.
8. Evaluate behavior cautiously. Test compatible tools, but do not infer universal adoption from one result.

Where llms.txt fits in GEO

Generative Engine Optimization is a broader discipline. It includes entity consistency, useful content, technical accessibility, accurate structured data, independent corroboration, and measurement across AI engines. For a tool that explicitly supports the proposal, llms.txt may provide a curated resource map. It cannot improve Google Search rankings, create OpenAI crawler access, or repair thin content, contradictory product facts, and weak authority.

For a complete strategy, combine this file with the seven-step GEO playbook and a repeatable AI search visibility baseline.

Frequently asked questions