What llms.txt is
llms.txt is a Markdown text file placed in the root of the domain, at /llms.txt. Its aim is to give language models a clean summary of the site: who the brand is, what it does and which pages are worth reading to understand it.
The idea comes from a real problem: web pages are full of menus, scripts and elements that get in the way of a model that only wants the content. A short, well-organised file saves it work.
It was proposed by Jeremy Howard, of Answer.AI, in September 2024. Since then many websites have adopted it, especially technical documentation sites.
What it looks like inside
The format is simple and designed to be read by both people and machines.
- A title (H1) with the name of the site or brand.
- A short summary, in a blockquote (>), with the essentials: what it is and who it is for.
- Sections (H2) with lists of links to the key pages, each with a sentence explaining what it contains.
- An optional section with secondary pages that the model can skip if it is short on context.
- Optionally, an llms-full.txt with the full content of the pages in Markdown.
Does it really work?
We have to be honest: llms.txt is not an official standard, and none of the big players (Google, OpenAI, Anthropic or Perplexity) has confirmed that it uses it to choose what to cite. Google has said publicly that it doesn't use it in search.
What we do know is that assistants read pages when they search in real time. A file that summarises the brand and links to the right pages does no harm and can help when an agent or a tool looks it up.
My recommendation: implement it because it takes an hour, but don't sell it as the answer to GEO. Anyone promising that an llms.txt will get you into ChatGPT is exaggerating.
How to implement it on your site
For an ecommerce site, the llms.txt should summarise the brand and link to what an assistant needs in order to recommend it: main categories, buying guides, shipping and returns, and the about page.
- Write the summary the way you would explain it in one sentence to a new customer.
- Link to categories and guides, not to hundreds of individual products.
- Include the policies: shipping, returns, warranty and contact.
- Publish it in the root of the domain and check that robots.txt doesn't block AI crawlers.
- Review it when the catalogue or the site structure changes.
What really moves GEO
llms.txt is a detail. What makes an AI cite a brand is the same thing that makes it trustworthy to a person: content that answers questions directly, your own data, consistent structured data, a fast site that can be read without relying on JavaScript, and mentions in the media, comparisons and third-party reviews.
If you have to choose where to put the hours, put them there. As for the llms.txt, get it done in an afternoon and move on.
Frequently asked questions
Is it mandatory to have an llms.txt?
No. It is a voluntary proposal, not an official standard. Not having one is not penalised by Google or by any assistant.
Does llms.txt replace robots.txt?
No. robots.txt tells crawlers what they can visit; llms.txt summarises the site for language models. They complement each other.
Where does the llms.txt file go?
In the root of the domain, so that it opens at yourdomain.com/llms.txt, just like robots.txt or sitemap.xml.
We’ll tell you what we would do in your case, no strings attached. A real person from the team replies.
Founder and Chairman of Webimpacto. He drives the adoption of AI across the agency and its clients’ projects: GEO, agentic commerce and new channels such as ChatGPT Ads.