llms.txt Explained: Do You Actually Need One?

If you've read any AI SEO advice in the past year, you've been told to add an llms.txt file to your site. It's presented as the robots.txt of the AI era: a quick file that tells ChatGPT and friends what to read. The pitch is appealing, since it takes twenty minutes and sounds like a secret handshake.

So does it work? The evidence so far says no, at least not for the thing most people want it for. Here's the full picture, so you can decide with your eyes open.

What llms.txt is

The llms.txt proposal, published in 2024 by Jeremy Howard of Answer.AI, is a plain Markdown file placed at the root of your site (yourdomain.com/llms.txt). It gives a language model a curated, readable summary of your site: what it is, and links to your most important pages, ideally with short descriptions. The idea is that a model trying to understand your site in limited context could read one clean file instead of wading through navigation, ads, and scripts. You can read the proposal at llmstxt.org.

A minimal example looks like this:

# Example Plumbing Co.

> Licensed residential plumbing in Springfield. Emergency repairs, water heaters, and drain cleaning.

## Services
- [Water heater repair](https://www.example.com/water-heaters/): Repair and replacement, same-day.
- [Drain cleaning](https://www.example.com/drains/): Clogs, camera inspection, and hydro-jetting.

## About
- [About us](https://www.example.com/about/): Licensing, service area, and team.

What the platforms have said

This is the important section, because the claim "llms.txt boosts AI visibility" is testable against what the companies themselves say.

• Google says you don't need it. Its guidance on generative AI search states that you don't need to create "new machine readable files, AI text files, markup, or Markdown" to appear in Google Search, including its AI features. That's in Google's AI optimization guide.
• Google's John Mueller was more pointed. On Google's Search Off the Record podcast, as reported by Search Engine Journal in June 2026, he argued that the file is essentially a site saying "I have the best website ever," and that LLM systems "by design, can't trust what is here as a way of differentiating between different websites."
• The other AI companies haven't committed. As far as we can find, OpenAI, Anthropic, and Perplexity have not published documentation saying their search or answer products read llms.txt to decide what to cite.

Mueller's argument is the keywords meta tag all over again. Search engines stopped trusting it long ago because anyone could put anything in it. A file you write about yourself has the same problem: it can describe you, but it can't convince a system you're better than the next site.

Where llms.txt might still be useful

"Not used by search engines" isn't the same as "useless." There are narrower cases where a clean, machine-friendly index is genuinely handy:

• Developer documentation. Coding assistants and agents that fetch docs can benefit from a tidy Markdown index. This is where the format took off, and many documentation platforms generate the file automatically.
• Your own AI tooling. If you're building a chatbot or agent that reads your site, an llms.txt is a convenient input you control.
• Costless side effects. Writing one forces you to decide what your ten most important pages are. That exercise is worth doing even if no bot ever reads the file.

What it doesn't do

• It doesn't grant or block access. Only robots.txt and your firewall do that. See Which AI Crawlers Should You Allow?
• It doesn't improve rankings in Google or Bing.
• It doesn't make an AI system cite you. No platform has documented that behavior.
• It isn't a substitute for a sitemap. A sitemap is read by real crawlers today.

Our recommendation

Don't make it a priority, and don't pay anyone to "implement" it as an AI visibility service. If your CMS or docs platform generates the file with one click, fine, turn it on. If you enjoy writing one, do it as a ten-minute prioritization exercise. Just don't let it displace the work that has evidence behind it.

That work, in rough order of return:

1. Make sure crawlers can reach your pages and see the content without running JavaScript.
2. Publish something worth citing: original information, clearly written. See How to Write Content AI Search Engines Cite.
3. Make your business unambiguous with accurate structured data and consistent profiles. See Schema Markup for AI Search.
4. Measure what happens, so you can tell whether any of it works. See How to Track Your Visibility in AI Search.

A caveat, since this field moves fast: if a major platform announces support for llms.txt, this advice changes. Until one does, the honest position is that it's harmless and unproven. For the bigger picture, head back to the AI Search Visibility guide.