Does llms.txt do anything for AI search?

You may have been told to add an llms.txt file to your site so that AI assistants will find you. Here is what the evidence shows, so you can decide in five minutes and move on.

The short answer

For getting named by AI search, there is no good evidence that llms.txt helps, and Google says its Search does not use it. If you want one for another reason, it is cheap to make. If you do not, you are not missing a lever.

What llms.txt is

It is a plain text file at yourdomain.com/llms.txt. The idea is a short, readable index of your most useful pages, written for AI tools. It is a proposed format. Ahrefs found that 28% of the sites it studied had one (see below). Unlike robots.txt, it does not allow or block anything. It is only a list.

What Google says

Google's guide to generative AI features in Search says you do not need to create new machine-readable files, AI text files, markup or Markdown to appear in Google Search, and that Google Search itself does not use them. It lists llms.txt among these. It says making one is optional and will neither harm nor help your visibility or rankings in Google Search, because Search ignores them. The page says it was last updated on 10 July 2026.

Google adds that its generative AI features are rooted in its core Search ranking and quality systems and highlight content from the Search index. So the usual basics still apply: pages that can be crawled and indexed, and helpful content.

Note what this covers. It covers Google Search. It does not tell us about every other AI product.

What the Ahrefs study found

Ahrefs published a log study on 15 June 2026. It looked at 137,210 websites that had traffic in May 2026, and checked who asked for their llms.txt file. Ahrefs says:

• 28% of the sites had the file. Ahrefs treats that as an upper limit, because its customers lean technical.
• 97% of those files got no requests at all in May 2026.
• 96% of the requests that did arrive came from bots, and 4% from people.
• Named AI tools made up 19.5% of requests. AI retrieval bots (the ones that fetch pages to answer questions) accounted for 233 requests, about 1.1% of the total.
• SEO audit tools accounted for 21.7% of requests.
• On sites with no file, nothing from AI bots came looking for one.

Ahrefs' reading is that the closest thing to an intended audience in its data is coding agents. For showing up in ChatGPT, Perplexity or AI Overviews, it calls an llms.txt file "largely decoration".

Limits of that study

Ahrefs lists some of these itself. It says a fetch does not mean the file was read, so every figure is a ceiling. Its customers skew technical and SEO-aware, so they are not typical small businesses. It says it did not check whether files followed the llms.txt specification. It measured one month, May 2026, so it says nothing about next year.

When it may still be worth having

• You publish documentation or tools for developers. Coding agents are the closest thing to an intended audience in Ahrefs' data.
• It takes you ten minutes. If your site tool makes it for free and you will not forget it, there is little downside.
• You want a tidy list for your own team. It can double as a map of your best pages.

If you add one, keep it short, link only to pages you already stand behind, and do not put anything in it you would not want read. A public file is public.

What to do instead

The things that decide whether an assistant can use your site are more ordinary.

1. Let the search crawlers in. OpenAI, for example, says sites that block its search crawler OAI-SearchBot will not appear in ChatGPT search answers. See Check that AI search bots can reach your website.
2. Say plainly who you are, what you sell and where, on pages anyone can open.
3. Keep your facts the same everywhere.

How to check it worked

Look in your server logs, or ask your host, for requests to /llms.txt over a month. If you see almost none from known AI bots, that matches the study. Ahrefs' advice is to check your own logs before investing further.

What this won't do

An llms.txt file will not make an assistant recommend you, and nobody can promise it will. It is not a security control either: it cannot block anyone. And the study may look different a year from now.