What is an llms.txt, and do I need one?

What is an llms.txt?

An llms.txt is a plain text file in the root of a website, at your-domain.com/llms.txt. It is written in Markdown and sums up for language models who is behind the website and which pages matter most, each with a link and a short sentence.

The idea comes from Jeremy Howard, who proposed it on llmstxt.org on 3 September 2024. The thinking behind it: a website is full of menus, banners and scripts that get in a language model’s way. A short overview in plain text helps it find the essentials quickly. The proposal also suggests offering important pages as a clean Markdown version.

llms.txt is a proposal, not a standard of a standards body. Whether and how an AI service reads it is up to each service.

What does an llms.txt look like?

According to llmstxt.org the file has a set order. Only the first part is required:

1. A heading with the name of the company or website, as an H1 (# Name). This is the only required part.
2. A short summary as a quote (> ...): who you are and what you offer.
3. Further details in plain text, such as location, prices or contact.
4. Sections with lists of links, each under an H2 heading (## ...): one link per line and, after a colon, a sentence about it.
5. An «Optional» section for pages a model can skip when it is short of room.

This is how the llms.txt of anewera.ch begins (shortened). We build it from the same data as the website, so it never goes out of date:

# anewera

> anewera is the Swiss AI visibility agency of Loaded AG. We measure what ChatGPT, Gemini, Google AI Overviews and Claude say about companies, and work to get them recommended.

anewera ist die Schweizer Agentur für KI-Sichtbarkeit der Loaded AG.

## English

- [What is GEO? Generative Engine Optimisation](https://anewera.ch/en/knowledge/what-is-geo): GEO explained: what Generative Engine Optimisation is, ...

## Optional

- ...

The whole file is at anewera.ch/llms.txt.

Do I need an llms.txt?

No, it is not required, but it costs little. Three points help you decide:

• It does nothing for Google. Google writes in its guide to generative AI features that Google Search does not use llms.txt files, not for AI Overviews and not for AI Mode either. An llms.txt for other services does no harm in Google Search, though.
• For other services it is open. We have found no statement by OpenAI or Anthropic that their assistants read the llms.txt of company websites. Nor have we found evidence from the providers that an llms.txt on its own raises the mentions in AI answers (as of 21 October 2026).
• For AI agents and developers it is useful. Programs that read websites on behalf of people find what they are looking for faster with an llms.txt. According to llmstxt.org the AI labs publish llms.txt files for their own developer documentation.

Our advice: if your website has the basics in place, an llms.txt is a small, sensible addition. If not, start with the basics further down.

Do ChatGPT, Claude and Gemini read llms.txt?

For company websites we have found no statement from the providers on this (as of 21 October 2026). What the providers do document is their crawlers and how to control them in robots.txt:

Provider	Crawler	What for (according to the provider)
OpenAI	OAI-SearchBot	showing websites in ChatGPT’s search results
OpenAI	GPTBot	content that may be used to train the models
OpenAI	ChatGPT-User	visiting pages when a person asks for them in ChatGPT
Anthropic	Claude-SearchBot	indexing pages for Claude’s search results
Anthropic	ClaudeBot	content that may be used for training
Anthropic	Claude-User	visiting pages when a person asks Claude

According to OpenAI the settings are independent: allowing OAI-SearchBot and blocking GPTBot lets you appear in ChatGPT’s search without releasing content for training. Anthropic writes that blocking Claude-SearchBot may make a site less visible in Claude’s search results. For Google’s AI Overviews and AI Mode, the normal Googlebot applies.

llms.txt, robots.txt and sitemap.xml: what is the difference?

All three sit in the root of your website, but they do different jobs:

File	For whom	Job
robots.txt	all crawlers	allows or forbids reading pages, per crawler
sitemap.xml	search engines	lists every page with the date it last changed
llms.txt	language models and AI agents	sums up who you are and names the most important pages with a sentence each

robots.txt is the only one of the three that forbids anything. An llms.txt is of no use if robots.txt locks the AI crawlers out.

How do I create an llms.txt?

In five steps:

1. Choose the pages: the ones that explain your offer best, such as services, prices, about us, contact and your best guides.
2. Write one sentence per page: what is on the page, factual, without sales talk. The page’s meta description is often a good start.
3. Build the file: H1 with the name, summary as a quote, sections with lists of links, «Optional» last.
4. Upload it: as llms.txt in the root, reachable at your-domain.com/llms.txt, as plain text.
5. Keep it current: new pages, changed prices and removed addresses belong in it straight away. The simplest way is to let your website system generate the file, as anewera.ch does.

Then check robots.txt: does it allow OAI-SearchBot, Claude-SearchBot and Googlebot?

What matters more than an llms.txt

An llms.txt replaces none of the basics on which AI services find and recommend a company:

• Open access: the AI services’ crawlers may read your pages, and your pages are indexed by Google.
• Clear pages: every important question your customers ask has a page that gives the answer in its first sentence.
• Sources: what directories, associations and trade media write about you is read by the AI too.
• Measurement: only continuous measurement of the answers, with a margin of error, shows whether a change works.

How this fits together is explained in AI SEO: SEO for ChatGPT and AI search. Whether your website is ready for AI agents can be checked for free at agentready.ch, like anewera a service of Loaded AG. If you want us to do the work, the llms.txt is part of Managed at the GEO agency anewera.