A proposed Markdown file placed at yourdomain.com/llms.txt that summarises a site and links to its most useful pages, written to be easy for language models and tools to read.

Do ChatGPT, Claude, Gemini or Perplexity use llms.txt?

Check each provider's current documentation. Do not assume it affects recommendations, and do not rely on it in place of clear pages and crawl access.

Can llms.txt hurt me?

It is unlikely to cause harm if it is accurate and kept in step with your site. An out-of-date file that points to old pages is a small risk.

Is llms.txt the same as robots.txt?

No. robots.txt tells crawlers what they may fetch. llms.txt is an optional summary that points readers to the content you consider most useful.

llms.txt explained: what it is and whether you need one: Citewake

llms.txt is a proposed convention: a plain Markdown file at the root of your site that gives a short description of what you do and links to your most useful pages, so that language models and tools can find them easily. It is optional, and it is not a substitute for clear pages that crawlers can read. Whether any given assistant uses it is something to verify, not assume.

If you can publish one in an hour and keep it accurate, it is cheap. It is not where most of the value lies.

What is llms.txt, exactly?

It is a proposal for a standard file placed at yourdomain.com/llms.txt. It is written in Markdown, so it is readable by both people and machines. A typical file contains:

a title and a one- or two-sentence summary of the site;
optional notes with key context;
sections of links, each with a short description, pointing to pages the owner considers most useful, such as documentation, pricing or guides.

It was put forward as a way to give models a clean, curated map of a site, without the clutter of navigation, scripts and adverts. Some documentation sites also publish plain-text versions of their pages for the same reason.

Is it an official standard?

No. It is a community proposal, not an internet standard with a formal body behind it. Unlike robots.txt, which has long-standing crawler support, there is no guarantee that any given AI product reads llms.txt. Support varies, and it changes, so check the current documentation of the assistants you care about instead of relying on a claim from a blog post, this one included.

What can be said safely: as far as public documentation shows, none of the major assistants has stated that recommendations depend on it. Treat it as optional.

How does it differ from robots.txt and a sitemap?

They do different jobs.

File	Purpose	Audience
robots.txt	Says which paths a crawler may fetch	Crawlers
sitemap.xml	Lists URLs so search engines can discover them	Search engines
llms.txt	Summarises the site and links to key pages, in readable form	Language models and tools

robots.txt is a permission. A sitemap is an inventory. llms.txt is a guided tour. For permissions, see AI crawlers and robots.txt.

Do you need one?

Probably not as a priority. Ask three questions.

Is your site hard for a machine to read? If your key text sits behind scripts, logins or tangled navigation, fix that first. A file pointing at pages a crawler cannot reach does not help.
Do you have documentation or reference content that people, or coding tools, fetch directly? An llms.txt is most plausibly useful there, because it gives tools a concise index.
Can you keep it accurate? A short, correct file is fine. A neglected file pointing at retired pages is clutter.

If the answer to the second is yes and the third is yes, it is a reasonable small addition. For most business sites, the effort is better spent on the basics: a clear homepage, honest pricing, pages that answer real questions, and presence on the third-party pages assistants cite.

What should go in the file if you write one?

Keep it short and factual. A useful structure:

# Your company name

> One or two sentences: what you do, who it is for.

## Key pages
- [Pricing](https://example.com/pricing): plans and what each includes
- [Product overview](https://example.com/product): what it does
- [Documentation](https://example.com/docs): setup and reference

## About
- [About us](https://example.com/about): who runs the company

Guidelines:

Describe pages in plain words. Do not try to persuade.
Link only to pages that are public and current.
Make sure the description matches the page. If the file says one thing and the page says another, you have created a conflict.
Review it whenever pricing, products or key URLs change.

What matters more than llms.txt?

Several things, all of which are more reliably linked to what assistants read.

A homepage that says plainly what you are. See why your homepage confuses AI assistants.
Crawl access for the search and user-triggered crawlers.
Structured data where it fits, covered in structured data and AI answers.
Pages that answer real buyer questions, in the first paragraph.
Presence on the third-party pages assistants cite. The kinds of pages AI assistants cite most shows what to look for.

How can you tell whether it made a difference?

Test, do not assume. Before adding the file, record how assistants answer your key questions, and which pages they cite. After adding it, repeat on the same questions after some weeks. Because answers vary run to run, only treat a change as real if it is larger than ordinary variation. The routine is in how to track AI visibility over time.

What are the risks of adding one?

The risks are small. A file that is inaccurate or out of date can send a reader to the wrong page, and a very long file defeats the point of a summary. Do not use it to make claims you would not put on your homepage, and do not stuff it with phrases. If you publish one, treat it as part of your site: owned, reviewed and corrected like any other page.

If you want to see how a site appears today, try the Citewake checker. Citewake's own measures are described in the methodology.