llms.txt explained: a table of contents of your site for AI
An llms.txt file tells AI systems in plain text what is on your site and where to find the most important pages. What it is, what it is not, and how to make one.
New to this topic? Start with What Is GEO? Generative Engine Optimization Explained
The problem llms.txt wants to solve
A website is built for people with a browser. Menus, buttons, cookie banners, scripts that only load content when you scroll. For a person that works. For an AI system that wants to understand quickly what a site is about, it is noise.
A language model answering a question about your field has limited room to read. If it first has to get through fifty lines of navigation and a cookie notice, there is less left for what you actually have to say.
llms.txt is a proposal to make that simpler: one text file in the root of your site that explains in a few paragraphs what you do and points to your most important pages.
What goes into an llms.txt
The proposal was published in September 2024 on llmstxt.org. The file is plain Markdown, the same simple formatting as this article, with a fixed structure:
| Part | What it contains |
|---|---|
| Title | The name of your business or site, as the main heading |
| Summary | A short paragraph saying what you do and for whom |
| Notes | Optional: what a reader needs to know to understand your site |
| Sections with links | Per topic a list of your key pages, each with a short description |
The file lives in a fixed place: yoursite.com/llms.txt. Not in a subfolder, because nobody looks there.
A simplified example for an installation company:
# Van Dijk Installation
> Installation company in Eindhoven for heat pumps,
> solar panels and maintenance for homeowners.
## Services
- [Heat pumps](https://vandijk.nl/heat-pumps): hybrid and fully electric
- [Maintenance](https://vandijk.nl/maintenance): yearly contract
## Knowledge
- [Subsidy 2026](https://vandijk.nl/subsidy): what you get back
What llms.txt is not
There are misunderstandings going around, so here are three things the file does not do.
It is not robots.txt. With robots.txt you give crawlers permission or forbid them something. llms.txt forbids nothing and allows nothing. It only points the way. If you want to decide which AI bots may read your site, you still do that in robots.txt.
It is not an official standard. It is a proposal that many sites adopt. Not all major AI providers have confirmed that they use the file. Anyone promising that an llms.txt gets you into ChatGPT is promising something nobody can guarantee.
It does not replace your content. A signpost to thin pages produces thin answers. The file only works when the pages it points to really answer your customers' questions.
Why you do it anyway
If it gives no guarantee, why spend time on it? Because the cost is small and the risk is zero.
An llms.txt takes an hour to make. It also forces a useful exercise: writing down in a few sentences what you do, and choosing which ten pages really matter. Many businesses discover along the way that their most important service has no page of its own.
And AI search engines are developing fast. What is optional today may be normal in a year. If the basics are already in place, there is nothing to catch up on.
First the basics that are proven
An llms.txt does nothing if AI crawlers are not allowed to read your site. So check your robots.txt first. Many sites unintentionally block exactly the bots that could cite them, for example through a security setting at the hosting provider or an old rule that keeps out all unknown bots.
Next comes structured data: an Organization schema that says who you are, and schema per page where it fits. Only once those two are in place is llms.txt the logical third step. How to make your content itself citable is covered in writing so AI cites you.
How to make one
- Write the summary. What you do, for whom and where. Two or three sentences.
- Choose your most important pages. Services, frequently asked questions, your best explainer articles. Better ten good ones than a hundred.
- Group them by topic and give each link a short one-line description.
- Put the file in the root and check that
yoursite.com/llms.txtreads as text in a browser. - Keep it up to date. When an important page is added or removed, update the file.
The dGENIX site has one itself, at dgenix.com/llms.txt. It updates automatically with every new blog post, so it never falls behind.
Common mistakes
- Including every page. A list of five hundred links is no longer a signpost. Choose what a new customer should read first.
- Marketing language in the summary. "The best partner for all your sustainable dreams" tells an AI system nothing. "Installation company in Eindhoven for heat pumps" does.
- Links to pages that no longer exist. A moved page without a redirect makes the file unreliable.
- Serving the file as HTML. It needs to be plain text. If your browser opens a styled page with a menu, the location or your server setting is wrong.
How dGENIX helps
The GEO Engine checks whether fourteen AI crawlers are allowed to read your site and which ones you keep out. After an audit it generates an llms.txt tailored to your site content, together with the schema markup AI needs to recognise you as a business. You place it yourself: the engine changes nothing on your site.
Want to understand what GEO is about first? Read what is GEO. A broader approach is in optimizing for AI search engines.
Frequently asked questions
Will an llms.txt get me into ChatGPT's answers?
Nobody can guarantee that. The file makes your site easier to understand for AI systems that read it. Whether you get mentioned depends mostly on the quality of your content and how often others mention you.
What is the difference with robots.txt?
Robots.txt decides which bots may read your site. llms.txt points the way to your most important content. You need both, and robots.txt comes first.
Should I also make an llms-full.txt?
The proposal mentions an extended variant with the full text of your pages in it. For most small businesses the regular llms.txt is enough. Start there.
