llms.txt is an optional community proposal for publishing a concise Markdown summary and selected public links at /llms.txt. The proposal is documented at llmstxt.org. It is not a W3C or IETF standard, a Google requirement, a robots directive, or a universal AI search standard.
No AI system is required to read, index, or cite the file. Publishing it does not guarantee discovery, answer accuracy, traffic, citations, or rankings. Its clearest benefit is editorial: a site owner can maintain a compact public briefing with links to important pages.
A conservative llms.txt example
# Acme Analytics
> Acme Analytics provides reporting software for small retail teams.
## Product
- [Product overview](https://example.com/product): Current features and supported workflows
- [Pricing](https://example.com/pricing): Current public plans and limits
## Help
- [Documentation](https://example.com/docs): Setup and usage guidance
- [Contact](https://example.com/contact): Public support optionsThe example follows the proposal's basic idea: a title, a short summary, and curated links. It does not contain instructions that override the linked pages or promises about how an external model will interpret them.
What belongs in the file
Choose pages that are public, current, and useful on their own:
- the canonical product or service overview;
- current pricing or plan limits;
- documentation and help;
- important policies;
- a small number of factual guides or reference pages.
There is no universal rule that every file must contain a particular number of links. Use the smallest set your team can keep accurate. A large site may need sections; a small site may need only a few links.
What to leave out
Do not publish:
- secrets, private URLs, customer data, or internal notes;
- claims that are absent from or contradicted by the linked pages;
- expired prices, deprecated features, or unsupported languages;
- commands that pretend to control third-party AI behavior;
- copied testimonials or statistics you cannot verify.
The file is public content and should pass the same evidence review as a landing page.
Deployment steps
- Draft the file from current public pages.
- Use absolute HTTPS URLs.
- Save it as plain UTF-8 Markdown at
/llms.txt. - Request the URL in a browser and confirm a successful response.
- Open every linked page.
- Add the file to your normal content review process.
You may link to /llms.txt from a human-facing page if you want people to find it. Adding an unofficial Llms-txt: line to robots.txt does not make discovery certain and should not be described as a recognized robots directive.
What about llms-full.txt?
The proposal also describes an optional llms-full.txt companion containing expanded Markdown content. It is not essential or required. A large duplicate file can become stale quickly, so publish it only if your team has a clear maintenance reason.
For many sites, accurate HTML pages plus a short llms.txt are easier to maintain than a second full-content copy. If two versions disagree, the duplicate creates ambiguity rather than clarity.
llms.txt and Google AI features
Google's current generative AI search guidance explicitly says Google Search does not use llms.txt. Google says publishing the file will neither harm nor help a site's visibility or rankings in Google Search, including its generative AI features. It may still be maintained for a different service or internal workflow that documents support, but it is not a Google visibility requirement.
Google also says ordinary SEO foundations remain applicable and that indexing and serving are not guaranteed even when a page meets its requirements and policies.
Keep the established foundations first:
- useful, accessible pages;
- crawlable links and accurate canonical URLs;
- appropriate indexing controls;
- current sitemap entries;
- structured data that matches visible page content where eligible;
- clear ownership and update processes.
How to measure the file honestly
There is no reliable universal effect data showing that llms.txt causes better rankings, citations, traffic, or answer accuracy. Start from a documented baseline and, where relevant, use server logs to check whether any known client requested the file. A request still does not prove that the file affected an answer.
Do not credit a visibility change to llms.txt merely because it happened after publication. AI answers vary, and other site or index changes may happen at the same time.
Instead, measure what the file itself can prove:
- it is publicly reachable;
- it parses as readable Markdown;
- every link works;
- its descriptions match the linked pages;
- the owner has a review date and update process.
If a specific AI product later documents support, test that documented behavior separately. Until then, treat retrieval observations as observations, not a universal causal result.
Common mistakes
Presenting the proposal as a standard
Call it an optional community proposal and link to its source. Do not imply endorsement by search engines or AI vendors.
Treating it as robots.txt
robots.txt communicates crawler access rules according to established crawler behavior. llms.txt is a content-format proposal; it does not grant access, force crawling, or override indexing controls.
Publishing unsupported marketing claims
Short files make claims easy to copy and repeat. That raises the importance of checking prices, capabilities, customer counts, comparisons, and outcomes before publication.
Letting the file drift
A stale summary can contradict the website. Assign an owner and review it whenever pricing, positioning, important URLs, or supported features change.
Frequently Asked Questions
Is llms.txt required for AI search visibility?
No. It is an optional community proposal. No AI or search system is required to read, index, or cite it.
Does llms.txt improve rankings or citations?
There is no reliable universal evidence that publishing the file causes a ranking or citation improvement. It can organize your own public information, but outcomes must not be promised.
How many links should llms.txt contain?
There is no universal number. Include a manageable set of current, important public pages. Accuracy and maintenance matter more than reaching a fixed count.
Is llms-full.txt required?
No. It is an optional companion described by the proposal. Publish it only when you can keep the duplicate content current.
Should llms.txt be listed in robots.txt?
You may choose to reference it, but Llms-txt: is not a standard robots directive and cannot ensure discovery. A normal public link is easier for people to understand and verify.
Can SkillAEO generate a starting file?
Yes. The Skills Pack generator can organize public information into a draft. Review every statement and link before deployment; generation does not prove accuracy or guarantee use by external systems.