The idea is simple: a plain text map of your site, written for machines, so an AI agent that visits can find your best pages without wading through menus and scripts. Whether any engine uses it is a separate question.
On that question the engines say little. Read this guide before you spend time on the file, and do the basics in our pillar guide first.
What is llms.txt?
According to the llmstxt.org proposal, authored by Jeremy Howard and published on 3 September 2024 (the page shows a modified date of 10 August 2026 and calls itself v2), llms.txt is a file that provides information to help agents use a website. It offers brief background, guidance and links to detailed Markdown files, and agents are expected to read it and follow the links they need.
The file is written in Markdown: an H1 with the project name, an optional short summary in a blockquote, optional detail sections, and H2 sections holding lists of links, each with an optional note. A section titled Optional holds links an agent can skip when it needs less context. The file can sit at the site root or at any path, where it covers the pages under that path.
What do the engines say about it?
Google says that to appear in its AI features you do not need to create new machine-readable files, AI text files or markup. That is as close as an engine comes to addressing a file like this, and it points away from needing one for Google.
We read OpenAI's crawler documentation and Perplexity's bots guide on 12 November 2026. Both explain how their crawlers behave and how to control them through robots.txt. Neither mentions llms.txt. The llmstxt.org page says AI labs publish llms.txt files for their own developer documentation, naming OpenAI, Anthropic and Gemini, but it does not describe any formal endorsement or support beyond publishing them. So no engine we read documents an effect on citations.
Who might benefit?
The proposal's own framing is about agents using a website, and its examples are documentation. Sites most likely to gain are those with large documentation, an API, or a product that developers use through AI coding tools, because an agent can read a clean map of those pages directly. For a typical local business or brochure site, the benefit is speculative: there are few pages to map, and ordinary crawling already finds them.
How do you write a sound llms.txt?
- Name the site in the H1 and add a one- or two-sentence summary of what it is and who it serves.
- List your key pages, each with a link and one line saying what it holds: product, pricing, documentation, policies, contact.
- Link to clean, current pages. Markdown versions are suggested for detail, but a link to a clear HTML page is a fair start.
- Use the Optional section for secondary material an agent can skip.
- Match the site. Every fact in the file must agree with the pages it points to.
- Put it at the root, for example your-domain/llms.txt, and keep a note of who owns updating it.
What are the risks?
The risks are small but real. A file that goes out of date will point agents to old facts. A file that lists material you meant to keep out of search can expose it, so never list private or gated pages. A file that contradicts the site, for example on prices or services, creates confusion. And a file can crowd out the more important work: clear pages, crawler access and accurate details. It is not a substitute for robots.txt, which is how OpenAI and Perplexity say their crawlers are controlled.
Should you publish one?
If your site has substantial documentation or an API, a short, accurate file is a cheap experiment. For most other sites, finish the basics first: let the engines' search crawlers reach your pages, answer questions near the top of each page, name authors and date facts. Our pillar guide on how to optimize for AI search covers those steps, and our technical audits check access and crawling.
Whatever you decide, do not buy llms.txt as a ranking tool. No one can promise a result from it, and the engines have not said it counts.
Conclusion
llms.txt is a proposal. Google says no AI text files are needed, and the other engine pages we read do not mention it. Treat it as an optional file, most plausible for documentation-heavy sites, and put clear, accessible pages first.
PromotEdge International is an international strategy, technology, marketing and design company, and our delivery team is in India. See our AI search service and our glossary entry on LLM SEO.
Facts we rely on: see our company facts for what we state about ourselves and where each fact comes from.