PL / EN Free consultation
Guide

llms.txt: what it is and does your business need one?

The llms.txt file shows up in nearly every GEO guide. Here is what it really is, what it does not do, and when adding it makes sense.

Author: Published: Updated: 6 min read

What llms.txt is

llms.txt is a 2024 idea proposed by Jeremy Howard of Answer.AI. The reasoning is simple: websites are written for people (menus, scripts, ads), while an AI model fits only a limited amount of text into its context window. A /llms.txt file in the root of a domain is meant to be a concise business card for the site in Markdown: a heading with the name, a short description and a list of the most important links with one-line descriptions.

Important: this is a community proposal, not a standard adopted by the W3C or IETF, and there is no guarantee that any specific bot will read it.

llms.txt vs robots.txt: not the same thing

Some guides describe llms.txt as “robots.txt for AI” or a consent file for model training. That is a misunderstanding.

robots.txtllms.txt
PurposeAllows or blocks bot access to URLsHints at which content matters most
StatusEstablished convention, respected by well-behaved botsProposal, no broad support
FormatUser-agent / Allow / Disallow rulesMarkdown with a description and links
Affects bot accessYesNo

If you want to control whether bots from OpenAI, Anthropic, Google or Perplexity may fetch your content, you do it in robots.txt by naming the relevant bots. Remember that different bots have different roles. Some collect data for training, others fetch pages when a user asks a question. Blocking all of them “to be safe” may stop AI from citing your site. Check each company’s documentation for current bot names and purposes, because they change.

Does llms.txt improve AI visibility?

The honest answer: there is no reliable evidence that it does. Google representatives have said outright that they do not use it in search, and the other companies have not described reading it when generating answers. Some developer tools and coding agents do use it, especially for technical documentation, and that is where it makes the most sense.

In our practice, whether a company is recommended is decided by other things: content that answers customer questions, clear information about who you are, structured data and mentions outside your site. That is how it went in the MOL case: results came from work on the site, content and web presence, not from one file.

When it is worth adding

  • You have documentation or a knowledge base (SaaS, API, e-commerce with extensive guides) and want to make life easier for tools that use it.
  • Your site is already in good order and you want to finish the technical details: it is about an hour of work.
  • You want a single, current summary of your offer that you can also hand over manually, for example in a brief or in a prompt for an AI tool.

If the site does not yet have solid content about the company and services, llms.txt will not fix anything. Start with the basics in 7 steps to get recommended by ChatGPT.

What a finished file looks like

The structure is simple. A shortened example (sample data):

# Company Name

> A short description in 1-2 sentences: what we do, for whom, where.

## Offer
- [Service A](https://example.com/service-a/): what it covers and who it is for
- [Service B](https://example.com/service-b/): what it covers and who it is for

## Knowledge
- [Pricing](https://example.com/pricing/): price ranges
- [Blog](https://example.com/blog/): guides and case studies

## Contact
- [Contact](https://example.com/contact/)

Write concretely, link only to pages you want recommended, and update the file when your offer changes. If you run several languages, prepare separate sections or separate files. Our site has such a file at /llms.txt, but we treat it as an extra, not as the main lever.

What to do instead of counting on one file

  1. Make sure bots can reach the pages you want promoted (robots.txt, no blocks at the server or CDN).
  2. Add structured data and a consistent company description.
  3. Publish content with answers to real customer questions.
  4. Earn mentions in independent sources.
  5. Regularly check what AI says about you.

Frequently asked questions

Does llms.txt affect Google rankings?

No. Google has said it does not use llms.txt, and the file is not a ranking factor. It does not replace robots.txt or structured data either.

Does llms.txt block AI bots?

No. Bot access is controlled by robots.txt. llms.txt only describes your site and points to important content, and it has no power to forbid anything.

Should a small service business add llms.txt?

You can, since it is an hour of work and costs nothing, but do not expect results. Content about the company, structured data and mentions in external sources have a much bigger effect on AI recommendations.

Kacper Bartoszak
Entrepreneur, owner of the PEOPLE HATE ADS* agency and SEO for AI. Has worked with brands such as Lech Poznań, Binance and Danone. About the author →