Should you add an llms.txt file to your website?

What is an llms.txt file supposed to do?

It is a plain text index at your domain root that points AI systems to the pages you want them to read. The idea is that a model summarizing your site reads the index instead of crawling every page and pulling it out of the HTML.

The format is markdown, saved as text at the root of the domain. A title, a line saying what the site is, then links grouped under headings with a short note under each about what the page covers. Anthropic publishes one for its developer documentation at platform.claude.com, and it reads like a table of contents a person could follow. Ours at corvan.io/llms.txt does the same for the pages we would want an assistant to quote.

Adoption is where it gets strange.

SE Ranking checked roughly 300,000 domains in a study Search Engine Journal wrote up on November 20, 2025, and found the file on 10.13 percent of them. The share was flat across low, mid, and high traffic tiers. Most visibility tactics show up first on the sites with the most to gain. This one landed everywhere at once.

Does having one get you cited more by AI tools?

Not as far as anyone has measured. The same SE Ranking study found no statistically significant correlation between having llms.txt and how often a domain gets cited.

The number that settles it for us is the one about the winners. Among the 50 most cited domains in that data, one had an llms.txt file. Forty nine did not, and those are the sites the models quote all day.

The domains getting cited are earning it in the pages they publish and the places that link to them. A text file at the root does not appear to add or subtract from that.

We get asked about llms.txt on calls, usually by someone who read a checklist that put it near the top. The study is the answer we give. If a text file moved citation frequency, some trace would show up across 300,000 domains, and none does.

Do OpenAI, Google, and Anthropic actually read llms.txt files?

Almost nothing reads it, and none of them have committed to using it. Limy's September 3, 2026 guide counted 500 million bot traffic events over 90 days and found 408 requests that targeted llms.txt directly.

408 requests out of 500 million events. Per that same guide, no major provider (OpenAI, Anthropic, Google, Meta, or Mistral) has publicly committed to using llms.txt as a ranking or retrieval signal. Google has been the most direct about it. John Mueller compared the file to the discredited keywords meta tag, as reported by Limy, and Google states it does not support the format.

The odd part is who publishes one anyway. Anthropic keeps a markdown index of its documentation at platform.claude.com, and it is one of the providers that has made no public commitment to using such a file as a signal. An index of your docs is useful to whoever wants a map of the site, which is a smaller claim than a ranking signal.

Is llms.txt worth the effort for a small team?

Yes, if you treat it as a five minute job and nothing more. We publish one at corvan.io/llms.txt and we do not count it as AI search work on any plan.

Ours took one sitting. We listed the pages we would want quoted, wrote a line under each, and saved the file at the root. When we add pages, we add lines. There is no maintenance story beyond that.

We are in an odd position on this one. We publish the file, we tell clients it will probably do nothing, and we would still be irritated to learn in a year that it quietly started mattering while we called it pointless. Both of those are true at once. Cheap, probably useless, and we did it anyway.