Reference

What is llms.txt, and do you need it?

The Brimm team · grounded in the docs

llms.txt is a plain Markdown file you place at the root of your site that lists and links your most important pages so a large language model can find them quickly. It was proposed in 2024, in the llms.txt proposal. Here is the honest part. A handful of AI tools chose to support it, but the biggest engine did not. Google said on the record: “Google does not use llms.txt and is not planning to.” So the file is optional and harmless, never a ranking factor and never required to be cited.

What the file actually is

llms.txt is not a standard. It is a proposal, published on September 3, 2024 by Jeremy Howard, co-founder of Answer.AI and fast.ai, at llmstxt.org. The idea is simple and not unreasonable: a single, clean, machine-readable index of the content you most want an AI model to read, written in Markdown so it parses without effort. You put it at /llms.txt, the same way robots.txt lives at the root.

The format is human-readable by design. It opens with an H1 site name, a short summary, then curated lists of links grouped under headings. A minimal version looks like this:

# Acme Roofing

> Roof repair and replacement in Austin, Texas. Free estimates.

## Core pages
- [Services](https://acme.example/services): what we fix and install
- [Service area](https://acme.example/areas): neighborhoods we cover
- [Pricing](https://acme.example/pricing): how estimates work

## About
- [Reviews](https://acme.example/reviews): verified customer reviews
- [Contact](https://acme.example/contact): phone, hours, address

That is the whole concept. A table of contents for machines. It does not change your pages, it does not block anything, and it does not feed the file's content into rankings. It is a pointer, nothing more.

Who reads it, and who does not

Adoption is mixed, and you should hear that plainly before you spend an afternoon on it. Some companies engaged with the idea early. Anthropic indicated support around late 2024, and Perplexity moved in 2025. So there are real tools that will look for the file. But two facts keep it from being a must-do.

First, adoption across the open web is low. In one large study, roughly 10% of domains had an llms.txt file at all, and the share of actual AI crawler traffic that requests the file is negligible. The crawlers that matter mostly are not asking for it.

Second, and this is the one that settles it for most owners, Google said no on the record. In July 2025 Google's Gary Illyes confirmed that Google does not use llms.txt and is not planning to. John Mueller, also at Google, compared it to the keywords meta tag, the long-discredited tag that sites once stuffed and search engines learned to ignore. When the largest source of AI answers and search traffic tells you a file does nothing for them, treat it as optional.

The honest verdict

llms.txt is harmless and optional. If you ship it, it might help a few AI tools that choose to read it find your best pages a little faster. That is a real, if small, upside, and there is no downside to a clean file. So if it takes you ten minutes, fine.

What it is not: it is not a Google ranking factor, and it is not a requirement for being cited in AI answers. We see owners treat it as a rule, as though a missing llms.txt is why they are invisible to ChatGPT or absent from Google's AI Overviews. It is not. Building the file will not move you into an answer you were not already eligible for. The folklore around this file is louder than the file deserves.

This matters because attention is the scarce resource. An hour spent perfecting a curated Markdown index that almost no crawler requests is an hour not spent on the work that actually decides whether an engine can quote you. We would rather be blunt about that than let a tidy-looking file stand in for the real job.

What to do instead

The work that pays off is the same work that has always paid off, and it has nothing to do with a special file. An AI engine can only quote a page it can reach and read. So spend the time there:

Let the right crawlers in. Check your robots.txt so you are not accidentally blocking the search and citation bots while trying to keep training bots out. Blocking the wrong line is the most common reason a site is never cited.
Serve your answer in real HTML. If your content only appears after JavaScript runs, a retrieval crawler sees an empty shell and quotes someone else. Load your page with JavaScript off and confirm your actual answer is there as text.
Be the most specific answer on the page. Lead with a self-contained answer near the top, include a real number or named entity, and link out to the source you cite. Vague pages get skipped.

None of that requires llms.txt. Google's own guidance on showing up in AI features says there is no separate framework for AI search: the same fundamentals apply, and the page has to be crawlable and readable first. Get those right and the file is, at most, a nice-to-have on top.

Check your own page

You do not have to guess whether any of this is in order. Paste your link into Brimm and we read your site the way the engines do: whether the citation crawlers can reach you, whether your answer survives without JavaScript, and how quotable your top passage is. We print the failures in fix order, and we will not tell you to go build an llms.txt to look busy.