llms.txt: what it is and is it worth creating one in 2026?

llms.txt is a markdown text file placed at the root of your site (/llms.txt) that summarizes what the site contains and points AI models to your most important pages. The idea is to hand AI a clean, curated version of your content instead of forcing it to parse HTML full of menus and scripts. In 2026 it is still a proposed convention, not an adopted standard: it sits on roughly 10% of sites and no major AI provider has confirmed using it in search.

It appeared in late 2024 and gained traction in the technical documentation world. But there is a lot of hype around it. This guide explains what it is, how to build one, and the honest truth about whether it does anything.

What is llms.txt?

It is a single markdown file, at the root of your domain, meant to be read by large language models. The problem it tries to solve is real: AI context windows are limited, and turning complex HTML pages into useful text is hard and imprecise. llms.txt offers, in one place, a concise description of the site plus links to the content that matters most. Think of it as an index written for machines.

How do you create an llms.txt file?

The format is deliberately simple and follows an order set by the spec at llmstxt.org:

• An H1 with the project or site name. This is the only required section.
• A blockquote with a short summary carrying the key information needed to understand the rest.
• Sections with headings (H2) grouping links to key pages, each link with a brief description of what it covers.
• Optionally, a final section with secondary resources the model can skip if context is short.

You create the file in a text editor, save it as llms.txt and place it at the root (reachable at yoursite.com/llms.txt). There are generators that build one from a URL, but the result is much better if you review the output by hand and tighten it around your canonical content.

Do AI engines actually read llms.txt?

Here is the part the hype hides: in AI search, almost nobody reads it. As of mid 2026, no major provider (OpenAI, Google, Anthropic, Meta, Mistral) has confirmed using llms.txt as a signal in their search or answer surfaces. In July 2025, Google stated explicitly, through Gary Illyes, that it does not support it and does not plan to. And server logs show that crawlers like GPTBot, ClaudeBot and PerplexityBot almost never request the file: they crawl the HTML directly.

Where llms.txt is actually consumed is a different context: developer tools and IDE agents (Cursor, Claude Code, Copilot, Windsurf), MCP servers and in-product AI assistants. That is why documentation sites like Anthropic, Vercel and Cloudflare already ship one, it helps anyone integrating their docs into developer tools, not anyone trying to show up more in ChatGPT.

llms.txt vs robots.txt: what is the difference?

They are different things despite the similar name. robots.txt is an established, respected standard that tells crawlers what they may or may not crawl. llms.txt is a proposal that tells models what the site contains and how to interpret it. One controls access; the other tries to give context. And unlike robots.txt, llms.txt still lacks broad adoption from the side that is supposed to read it.

Who is llms.txt actually for?

It makes the most sense if you have technical documentation, an API or a knowledge base you want well consumed by AI developer tools. In those cases a curated llms.txt improves how those agents cite and use your docs. For a marketing site, a blog or an e-commerce store chasing visibility in ChatGPT or Google AI Overviews, the return today is close to zero.

Is it worth creating one in 2026?

Honest answer: it is optional and low priority. Creating an llms.txt is harmless, cheap and signals intent, so if you have technical docs or want to get ahead of the curve, go for it. But do not expect AI citation gains: tests across more than 300,000 domains showed no correlation with more search visibility. The effort pays off far more on the factors that actually matter: brand mentions across the web, factual density and answer-first formatting. Build the llms.txt after you have those basics in place, not before.

Next step: instead of investing in llms.txt, focus on what moves the needle. Go deeper into the GEO pillar and learn, in the how to rank on ChatGPT guide, the factors that actually make AI cite you.