What is llms.txt — A Site Guide for AI
/llms.txt is a proposed specification that provides AI crawlers and LLMs with a site overview and links to key content in Markdown format. Think of it as an AI-oriented counterpart to robots.txt — designed to help AI better understand your site.
Placed at /llms.txt in the site root, it is intended to function as a guide page that AI crawlers can reference first when visiting your site. A companion file /llms-full.txt containing more detailed content is also proposed as an extended variant.
Minimal llms.txt Sample
# Site Name > Describe your site in 1–2 sentences. ## Docs - [Page Title](https://example.com/page): Brief description ## Optional - [Supplementary Page](https://example.com/sub)
Elements to include:
- Use H1 (#) for your site name
- Use a blockquote (>) to summarize your entire site in 1–2 sentences
- Organize links to key content under ## sections
- Add a brief description after a colon for each link
3 Use Cases for llms.txt
Present your site's key points to AI in GEO/AI search contexts
When AI search engines like ChatGPT or Perplexity crawl your site, having an llms.txt helps them efficiently understand your site's overview and key content. It serves as an experimental early-adoption measure to increase the chance that your site is included as a citation target.
Explicitly guide AI to your main content
By listing links to your important pages by section in llms.txt, you can guide AI on which content to prioritize. Larger sites especially benefit from making key pages less likely to be overlooked.
Supplement AI comprehension alongside structured data
While JSON-LD structured data targets Google's crawler, llms.txt plays a complementary role in describing your site in natural language for AI crawlers. Maintaining both improves the comprehensiveness of information available to crawlers.
An honest note: llms.txt is an unofficial proposed specification still in the advocacy stage. Major AI providers such as Google and OpenAI have not officially stated that they reference it. Adopting it is an experimental early-adoption step, and no guaranteed effect on search rankings or AI citations should be expected. Consider it a low-risk, forward-looking measure.