Technical guide · Ambassador Digital
The complete llms.txt template & guide for modern web apps
An llms.txt file is a proposed Markdown reading map at your website’s root. It gives AI tools a short description of your site and links to useful public content. It is not an access-control mechanism, a replacement for robots.txt or a guarantee of citations. Support varies, so treat it as an optional complement to accessible HTML.
Start with a small, accurate reading map
The llms.txt proposal defines a Markdown format with one required H1 site name. A short blockquote summary, explanatory paragraphs and H2 sections of links are optional. Keep the structure flat; put lower-priority reading in a final “Optional” section.
There is no universal token budget or required file length. Prefer a concise index of useful pages over duplicating your entire website. Recheck it when pages move or your service changes.
# Your site name
> A concise description of your product and audience.
Describe your site using verified facts.
## Pages
- [Home](/): Product overview.
- [About](/about): Organization and approach.
## Optional
- [Public guide](/guides/example): Supplementary reading.Replace the sample names and links, and remove any page that does not exist.
Select public pages, not internal surfaces
Include your service overview, public documentation, useful guides and public product information. Use one or two representative entries for public dynamic content rather than exporting every record.
- Exclude sign-in, signup, reset-password and verification pages.
- Exclude admin, staff, internal operations and role-gated pages.
- Exclude private account settings, billing and user dashboards.
- Exclude API, webhook and callback endpoints.
- Never include secrets, private documents or draft research.
A URL being reachable does not make it suitable for a public reading map. When in doubt, leave it out.
Publish it at the site root
- Save the customized text as public/llms.txt in a modern app with a public assets directory.
- Publish the site, then open /llms.txt on your public domain.
- Confirm the response is plain text, not a login page or HTML fallback, and that each linked public page opens without a session.
- Check crawler preferences and security rules separately. llms.txt does not override them.
- Update links and descriptions when significant public content changes.
Ambassador Digital’s own public reading map is a working example. It deliberately omits unpublished benchmark studies and does not make claims about universal AI support.
Do not confuse documentation with discovery guarantees
A reading map cannot repair inaccessible HTML, missing content, a crawler block or inconsistent brand facts. Maintain useful public pages, internal links and accurate structured data first. Measure any observed benefit instead of assuming that adding the file changes an engine’s behavior.
Common questions
Does llms.txt make ChatGPT cite my website?
No. The format is a proposal and support is not standardized. A file can offer a useful reading map to tools that consume it, but it does not guarantee discovery, crawling or citations.
How is llms.txt different from robots.txt?
robots.txt communicates crawler access preferences. llms.txt describes public content and suggests useful reading. Neither file should be used to protect private information; enforce access restrictions on the site itself.
Should I include account, admin or API pages?
No. Include only public informational pages intended for unauthenticated visitors. Omit login flows, admin surfaces, private dashboards, per-user pages, API endpoints and callbacks.
Do I need llms-full.txt?
Not for this starter setup. A longer file can package substantial public documentation for tools that support it, but it is optional and should not duplicate private or unverified material.