llms.txt explained: what it is and how to set it up

llms.txt is a small text file that lives in the root of your website. It has one purpose: to give AI models a clear map of your most important content, in a format they can read easily. Sounds useful, and it can be. But there is a big ‘however’ that I don’t want to keep from you.

In this article I explain what llms.txt is, what a good file looks like, and I give you a working example. Then I’m honest about the question that really matters: does it work yet in 2026, and should you spend time on it now? In the audits I run I get that question more and more, so let me answer it properly here once and for all.

What exactly is llms.txt?

llms.txt is a 2024 proposal from Jeremy Howard, co-founder of Answer.AI. The idea is simple. You place a file at yourdomain.com/llms.txt containing a short, Markdown-formatted list of your most important pages. That way an AI model doesn’t have to crawl your entire site, it gets the core handed to it right away.

Compare it to a table of contents you write specifically for AI. You say: these are my most important sources, this is what they are about, and this is how to interpret them.

Common misconception: llms.txt is not robots.txt

I often hear this wrong, so let me set it straight immediately. llms.txt is not about permission. It does not tell crawlers what they may or may not use. You handle that with robots.txt and with your settings for AI bots.

The difference in one sentence: robots.txt says ‘you may or may not crawl this’, llms.txt says ‘this is my most important content, and this is how it fits together’. One is about access, the other about curation.

What does an llms.txt look like?

The format is deliberately simple. The specification prescribes a fixed structure:

• An H1 with the name of your site or project. This is the only required part.
• A blockquote (a line starting with >) with a short summary of what your site does.
• Optionally one or more paragraphs with extra context on how the links should be interpreted.
• One or more H2 sections with links to your most important pages below them, each with a short description.
• A separate H2 named ‘Optional’ for less important links, which a model may skip if it wants to keep things short.

A working example

This is what a simple llms.txt for JSResults could look like. Paste this content into a file named llms.txt and place it in the root of your site, so it is reachable at yourdomain.com/llms.txt.

# JSResults

> JSResults helps brands become more visible in search engines and in
> AI search systems such as ChatGPT, Perplexity and Google AI Overviews.

JSResults is the freelance label of Julian Smits, specialist in SEO and GEO.

## Key pages

- [Services](https://jsresults.com/en/services/): overview of SEO, GEO and AI search optimization.
- [About Julian](https://jsresults.com/en/about/): background and expertise of the author.
- [Cases](https://jsresults.com/en/cases/): examples of results achieved.

## Blog

- [How ChatGPT picks sources](https://jsresults.com/en/blog/how-chatgpt-picks-sources): how AI selects sources.

## Optional

- [Contact](https://jsresults.com/en/contact/)
- [Privacy](https://jsresults.com/en/privacy/)

Does it actually work? The honest state of play in 2026

Here comes the ‘however’ from the intro. Because however neat the idea is, in 2026 the practice is still thin. I think you should know that before you spend an afternoon on it.

• The major AI crawlers barely fetch the file. GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot and Google-Extended almost always just crawl your HTML, not your llms.txt.
• None of the major players (OpenAI, Anthropic, Perplexity) has committed to reading the file automatically.
• It is not an official standard. There is no W3C or IETF behind it, and there is no enforcement.
• Adoption is low. A study by SE Ranking across 300,000 domains found adoption of around ten percent.

And Google? John Mueller of Google compared llms.txt to the keywords meta tag that nobody uses anymore. The Google guide on generative AI from May 2026 even states explicitly that machine-readable files such as llms.txt are not needed to appear in AI search results.

So should you make one?

My sober advice: don’t see llms.txt as a ranking factor and don’t expect short-term visibility from it. But the effort is small and the risk is zero. So this is how I would weigh it.

Do it if: you have a documentation-heavy or knowledge-rich site, or if a file is ready in a few minutes via a tool. Then you tick a box for the future.

Don’t spend time on it if: you still have low-hanging fruit that does work right now, such as structuring your content, indexing your site well in Bing, and building your authority. That delivers more today.

In other words: an llms.txt is a fine ‘nice to have’, not a priority. I advise clients to first get the basics right that do have a demonstrable effect. Which ones those are in your case, I map out in a GEO audit.

How do you set it up?

• Create a text file named llms.txt, in plain text with Markdown formatting.
• Fill it according to the format above: H1, blockquote, and H2 sections with your most important links.
• Place the file in the root of your site, so it is reachable at yourdomain.com/llms.txt.
• On WordPress? Yoast SEO can generate and update an llms.txt for you automatically. That saves maintenance.
• Check the file in your browser and keep it short. Only point to your genuinely important pages.

Frequently asked questions

What is llms.txt?

llms.txt is a text file in the root of your site that gives AI models a Markdown overview of your most important content. It was proposed in 2024 by Jeremy Howard of Answer.AI.

Do ChatGPT and other AIs use llms.txt yet?

Barely. In 2026 the major AI crawlers hardly fetch the file, and OpenAI, Anthropic and Perplexity have not committed to reading it automatically. They crawl your HTML directly.

Is llms.txt the same as robots.txt?

No. robots.txt governs access, meaning what a crawler may or may not fetch. llms.txt is curation: it points to your most important content. It does not grant or deny permission.

Should I create llms.txt for better AI visibility?

It is low effort with no risk, but not a ranking factor. Spend your time first on content structure, indexing in Bing and authority. Those have a demonstrable effect right now.

Sources

This article combines my own practice with publicly verified information. I checked the facts on format and adoption against:

• Ahrefs, What Is llms.txt, and Should You Care About It: ahrefs.com/blog/what-is-llms-txt
• Semrush, What Is LLMs.txt and Should You Use It: semrush.com/blog/llms-txt
• SE Ranking adoption study (cited in multiple industry sources, 2026).
• Statements by John Mueller (Google) and the Google guide on generative AI, May 2026.