← Blog

llms.txt for small business sites: what we know and don't

llms.txt is a proposed plain text file that points AI tools at your best pages. We have not measured any effect on rankings or AI answers, and here is what that means for you.

llms.txt is a proposed plain text file, placed at the root of a website, that lists your most useful pages with short descriptions so AI tools can read them quickly. That is all we can say with confidence. We have not measured any effect from it on search rankings or on whether AI assistants mention a business, and we are not aware of reliable evidence from anyone else. It gets searched for about 590 times a month in Australia, so people clearly want to know. Here is the honest version.

What the file is

The idea is a Markdown file at /llms.txt. It has the site name, a one line summary, and links to important pages, each with a sentence about what is there. Some sites also publish a longer llms-full.txt that contains the page text itself. The goal is to give a language model a clean, small map instead of making it work through navigation menus, scripts and ads.

It is a proposal from the community. It is not a web standard like robots.txt or sitemaps, and no search company is obliged to read it.

What we know

  • The format is simple and takes minutes to write.
  • Writing one forces you to decide which pages matter, which is a useful exercise on its own.
  • Serving it does not block or expose anything. It is just a file.

What we do not know

  • Whether Google, Bing, OpenAI or any other company's crawlers use it when building answers. We have no confirmation from any of them.
  • Whether it changes which businesses get named. We have not run a test that could show this, and a single site's before-and-after would not prove it either way.
  • Whether it will matter next year. The situation may change.

We should also say what we have not done. As at 13 October 2026, pagecog.com.au does not serve an llms.txt file. We have not run a controlled comparison on any site. So anything we told you about its effect would be a guess.

How we would test it, and why that is hard

A fair test needs a log of requests to the file, to see whether any AI crawler fetches it at all. That is the first thing to check, and it is free: search your server or Cloudflare logs for requests to /llms.txt. If nothing asks for it, it cannot be influencing anything.

Measuring a downstream effect is much harder. Rankings move for many reasons at once. We learned this with another change. In what we measured after adding answer blocks, the number that mattered was whether a page had been recrawled, and a quarter of the pages had not been when we looked. A simple before-and-after comparison is easily contaminated by other changes made in the same weeks. An llms.txt test would face the same problem, and AI answers vary from one request to the next as well.

Why the idea exists

Language models read web pages with limited context, and a typical page is mostly navigation, scripts and repeated layout. A short curated file looks like an obvious help. The weakness is that a search company already has a crawler, an index and a ranking system for deciding which pages are useful, and a file that a site writes about itself is the sort of signal that is easy to abuse. That is a reasonable explanation for why adoption by the big engines is unconfirmed, though it is our reading, not a statement from any of them.

A reasonable file, if you add one

Keep it short and honest. A site name, one plain sentence on what the business does and where, then links to your main service pages, pricing, contact and the handful of guides you most want quoted, each with a single descriptive line. Do not stuff it with keywords or claims you would not put on the page itself. Anything in the file should also be visible on the site, so the two never disagree.

What we would do instead

If your goal is to be named by an AI assistant, these have better reasoning behind them than a text file:

  1. Write answer blocks. A heading close to the question people ask, a one or two sentence direct answer, then two or three lines of support. They are easy for any system to lift out.
  2. Make sure the pages are indexed in Bing and Brave, which are reported to be used by some assistants. See setting up Bing Webmaster Tools.
  3. Do not block AI crawlers by accident in robots.txt, and do not hide content behind scripts that need to run before text appears.
  4. Keep your business details identical everywhere.

The full set is in how to get your business recommended by ChatGPT and other AI answers.

Should you add one?

If you have a developer and ten spare minutes, fine. List your services, pricing, contact and a handful of key posts. Do not pay for it as a service, and be wary of anyone who promises it will lift your visibility. Nobody can promise that.

If you want help working out what is worth doing for your own site, our SEO service starts with the basics, the pricing page shows what that costs, and get started is the way to ask. Related: IndexNow explained.

Common questions

What is llms.txt?

It is a proposed text file, placed at the root of a website, that lists your key pages with short descriptions in a simple format so AI tools can find them. It is a suggestion from the web community, not an official standard.

Does llms.txt improve my ranking in Google or ChatGPT?

We have not measured any effect and we know of no reliable evidence that it does. Do not expect it to change rankings.

Do ChatGPT, Google or Bing read llms.txt?

We cannot confirm it. Search companies have not committed to using it, and some have indicated they do not. You can check your own server logs to see whether any AI crawler requests the file.

Is llms.txt the same as robots.txt?

No. Robots.txt tells crawlers what they may fetch. llms.txt is a guide to what is most useful, and nothing in it blocks or allows access.

Should a small business add llms.txt?

It takes minutes and does little harm, so it is a low priority option. Clear pages, direct answers and being in Bing and Brave indexes matter more.