What is llms.txt, and should your business have one?

What is llms.txt?

llms.txt is a small text file that usually sits at the root of a website, like robots.txt or sitemap.xml. (The proposal also allows one in a subfolder, such as /docs/llms.txt, covering just that section.) Where robots.txt tells crawlers which URLs they may access and a sitemap lists the pages you want search engines to find, llms.txt is meant to be a short, readable briefing for AI tools: here is who we are, and here are the pages worth reading.

The idea comes from a proposal published at llmstxt.org by Jeremy Howard, dated September 3, 2024. The proposal's reasoning is practical. Web pages are built for people and wrapped in navigation, ads and scripts, which makes them hard to convert back into clean text for a language model. A concise file with links to clean versions of key pages saves the model work.

The proposal was revised as version 2 in August 2026. The update says thousands of sites now publish the file and that documentation platforms generate one automatically. That's the author's own description, not an independent count. Version 2 also added a way for pages to point to their llms.txt file using standard link tags or an HTTP header.

What llms.txt is not

A lot of marketing has grown up around this file, so it's worth being clear about its limits.

• It isn't a standard from a standards body. It's a proposal from an individual researcher that people have chosen to adopt. No search engine is bound by it.
• It isn't a ranking signal anyone has confirmed. As of October 2026, we found no statement from Google, OpenAI or Perplexity that they use llms.txt to decide which local businesses appear in answers.
• It isn't a permissions file. If you want to allow or block crawlers, that's still robots.txt. llms.txt doesn't control access.
• It isn't a replacement for good pages. The file only points to content. If the content is thin, a tidy index of it doesn't help.

Google is the most explicit. Its documentation on AI features says owners have no need for a new machine-readable file, an AI text file or extra markup for their pages to be eligible in Google's AI Overviews and AI Mode, and adds that no special schema.org structured data is required either. That doesn't forbid llms.txt. It does mean Google isn't asking for it. What the same page does list is the usual groundwork: allow crawling in robots.txt, link your pages to each other, keep important content in text form, and keep your Business Profile information up to date.

Who actually uses it?

The clearest real-world use is software documentation. The proposal says the file is used most heavily there, by coding agents that follow it to find API references and tutorials. The proposal's homepage also notes that OpenAI, Anthropic and Gemini publish llms.txt files for their own developer documentation. Publishing a file for your own docs is different from using other people's files to rank search results, and we found no statement from any of those companies that their consumer assistants rely on it to choose businesses.

There is one more sign of interest. Chrome's Lighthouse tool includes an llms.txt audit as part of its agentic browsing checks. Read the details, though. The audit flags a page only when requesting the file causes a server error. If the file simply isn't there, the audit is marked not applicable, and the documentation says providing the file is optional at the moment. The same page says that without the file, agents may spend more time crawling a site to work out its structure, which is a fair description of the benefit: a faster read, not a higher ranking.

How does it compare to robots.txt and a sitemap?

Three files can sit at the root of your site, and they're easy to mix up. Each does a different job.

If you only have time for one of these, get robots.txt right first. A single Disallow: / line can block a crawler from your whole site, and no summary file will make up for it. Block OpenAI's OAI-SearchBot crawler, for example, and OpenAI says your pages won't feature in ChatGPT's search answers, though a bare navigational link may still show up. Likewise, Perplexity's crawler documentation tells site owners to let PerplexityBot in so their pages can surface in Perplexity results. Google's own robots.txt introduction adds a warning worth knowing: robots.txt controls crawling, but it is not a way to keep a page out of Google.

What goes in the file?

The proposal defines a simple structure, in this order:

1. An H1 heading (a line starting with one #) with the name of the site or project. This is the only required section.
2. A blockquote (a line starting with >) with a short summary containing the key facts needed to understand the rest.
3. Optional paragraphs or lists with more detail about the site and how to read the linked files. Any Markdown except headings is allowed here.
4. Optional sections, each starting with an H2 heading (##), containing lists of links. Each list item is a Markdown link, optionally followed by a colon and notes about that page.

By convention, a section titled "Optional" holds secondary links an agent can skip when it needs a shorter read. The version 2 change notes say that section remains a useful convention but no longer carries any built-in processing meaning. The proposal also suggests offering clean Markdown copies of important pages at the same address, either with .md added to the end or with the file extension swapped for .md, so an agent can read the content without the page layout.

An example for a contractor

Here is what a file might look like for a fictional company. Every line is plain text.

• # Example Roofing Co.
• > Residential roof repair and replacement, gutters and storm damage inspections in Springfield and surrounding towns. Insured. Free written estimates. Phone (555) 010-2040.
• We work on asphalt shingle and metal roofs on homes. We do not do commercial flat roofs. Emergency tarping is available after storms.
• ## Services
• - [Roof replacement](https://example.com/roof-replacement/): what is included, timeline, warranty
• - [Roof repair](https://example.com/roof-repair/): leaks, missing shingles, flashing
• - [Storm damage inspection](https://example.com/storm-damage/): how inspections and insurance photos work
• ## Service area
• - [Towns we serve](https://example.com/service-area/)
• ## Company
• - [About us](https://example.com/about/)
• - [Reviews](https://example.com/reviews/)
• - [FAQ](https://example.com/faq/)

Notice what it contains: plain facts an assistant could repeat accurately. What you do, what you don't do, where, and how to reach you. Leave out slogans. A model reading "we're the most trusted roofer in the state" learns nothing it can use, and you've made a claim you can't back up.

How to add llms.txt to your site

1. Write the summary first. Two or three sentences covering services, area and contact. If you can't write this clearly, fix your website copy before anything else.
2. Pick ten to thirty links. Service pages, the service-area page, about, reviews, FAQ, and any guides customers actually read. Skip login pages, tag archives and thank-you pages.
3. Save it as llms.txt in UTF-8 plain text and upload it to the root of your domain, so it loads at yourdomain.com/llms.txt.
4. Check it loads in a browser without errors and without a redirect to a login or error page.
5. Keep it current. When you add a service or drop a town, update the file at the same time as the site. A stale file is worse than none, because it repeats outdated facts.

On most hosting this takes minutes. Some hosted website builders limit which files you can place at the root of your domain. If yours does, that's not worth a migration on its own.

Where does llms.txt fit in your plan?

Think of it as the last item on the list, after everything that's known to matter. The companion guide on getting named in ChatGPT answers covers the crawlers and settings that decide whether assistants can read your site at all. Our local SEO guide covers the profile, pages and reviews that feed both Google and the assistants.

We include an llms.txt file, along with a sitemap, robots.txt and structured data, on the sites we build, because it costs almost nothing and might help. We don't sell it as a lead source, and you shouldn't buy it as one. If someone quotes you a monthly AI search fee and this file is most of the work, ask what else is included. You can see how we handle it on our website builds, or get a free system quote that covers your site's search setup.

Frequently asked questions