What is llms.txt and why did we add it to our site?

Our site now has a small text file at dknmedia.gr/llms.txt. Does anyone read it? Does it bring us customers from ChatGPT? There are big promises around this file, so here is what it is, what the evidence says, even where it doesn’t suit us, and why we added it anyway.

The short answer

llms.txt is a short, organised description of your site, written for AI assistants. So far there is no evidence that it improves your ranking or earns you more mentions in AI answers. We added it because it costs very little and gives correct, collected information to any tool that asks for it, not because we expect miracles.

What llms.txt is

Jeremy Howard proposed it in September 2024. The problem it tries to solve is simple: an AI model reading a site to help someone has limited room, and a normal page is full of menus, scripts and ads. llms.txt gives it a clean starting point: what this site is and where to find the important parts.

Two things are worth remembering:

It’s a proposal, not an official standard. No company is obliged to read it.
It’s meant mostly for the moment of use, when an assistant looks something up on someone’s behalf, rather than for training models.

What it looks like

It’s a Markdown file at the root of the site. Only the title is required. Usually a short summary follows, then lists of links grouped into sections. A section called “Optional” marks what can be skipped when room runs out. A piece of ours:

# DKN Media

> Web agency in Paleo Faliro, Athens, Greece, run by developer
> Konstantinos Droulias. Builds websites, e-shops, custom web and
> mobile apps, and business systems.

## Services

- [Websites](https://dknmedia.gr/en/website-from-300/): from €300 + VAT
- [E-shops](https://dknmedia.gr/en/eshop-development/): from €600 + VAT

The proposal also describes clean Markdown versions of each page. We didn’t make those: for a site like ours, it would be effort with no proven gain.

How it differs from robots.txt and a sitemap

File	What it answers	Is it respected?
`robots.txt`	Where bots are allowed to go	Established, followed by major search engines
`sitemap.xml`	What all our pages are	Established, read by search engines
`llms.txt`	What matters and what it means	Optional, few ask for it

llms.txt replaces neither. It doesn’t stop anyone reading your site, and it doesn’t protect your content from model training. That’s the job of robots.txt.

Does it make you more visible in AI search?

Here the evidence is more disappointing than the ads. Before you pay anyone for this, look at what it says:

Google is explicit. Its guide to AI Overviews and AI Mode says: “You don’t need to create new machine readable files, AI text files”. Normal SEO is enough. Google’s John Mueller, according to Ahrefs, described it as something not done for search.
No link with mentions. SE Ranking looked at about 300,000 domains and found no correlation between the file and how often AI models cite a site.
Most files are never requested. Ahrefs found that about 97% of some 38,000 files received not one request in a whole month (May 2026).
Not even by bots looking for answers. In the same study, retrieval bots made just 1.1% of requests. SEO tools made more, at 21.7%.

The studies have limits: SEO tool companies ran them on their own data, and correlation proves nothing final. The field moves fast, and a company could start using the file tomorrow. But today, anyone promising you a rise in ChatGPT thanks to this file is promising something nobody has shown.

Who actually uses it

Its real use is elsewhere: in software documentation. Companies such as Stripe, Cloudflare and Anthropic have llms.txt on their docs, because developers ask AI assistants to read those instructions and write code.

The Ahrefs study shows it too. AI agents, meaning tools that carry out tasks for someone, were the largest AI group at 10.5% of requests. So when someone hands an assistant your site and says “find out what they do and what it costs”, a good llms.txt can get it to the right answer quickly. It can, not that it certainly will.

Why we added it, then

The cost is close to zero. One file, a few hours of work, no weight on the page. No visitor sees it and it slows nothing down.
We want our information to be right. A tool that asks for it finds the address, services and prices in one place, instead of guessing from the menu.
Our site is mainly Greek. The file is in English and points to both languages, which helps tools that work mostly in English.
We try things on ourselves first. We’ll check the server logs to see whether anyone requests it, and who. We prefer to talk to clients with our own evidence.

What we don’t expect: a better Google position, or more mentions just because it exists. Since we’ve only just added it, we have no results of our own yet. When we do, we’ll write them up.

What we put in and what we left out

The file isn’t a copy of the site. It’s a selection:

Who we are, where we are and how to reach us.
Our services, with the prices we already publish.
Our projects and articles. These update by themselves when we publish something, so the file doesn’t fall behind.
The Greek pages, under “Optional”.

We put in nothing that isn’t already public, and no promises or keywords. Anyone can read the file, so it’s no place for hidden information, and stuffing it impresses no model.

What really helps in AI search

If your goal is to appear in AI answers, these are worth far more, and they help normal SEO as well:

Let the bots in. Check your robots.txt. OAI-SearchBot, for example, decides whether you appear in ChatGPT search. GPTBot is separate and concerns training, so you can block one and not the other.
Put the important things in text. Prices and services that exist only inside an image are hard to read.
Answer your customers’ questions, clearly and from your own experience.
Link your pages properly, so they’re easy to find.
Keep your Google business profile up to date, if you have one, with the same details everywhere.
Build the fast, correct site you’d want to use yourself. For the numbers, see our article on Core Web Vitals.

Should you add one too?

You have documentation, an API or a developer product: yes, this is where it has real use.
You have a local business or small e-shop: it’s optional. It won’t hurt, but add it after everything above.
Someone is selling it to you as paid “AI SEO”: ask what evidence they have. It’s a few hours of work at most, not a service worth a subscription.

How to make one in 30 minutes

Write a title with your name and two or three lines on what you do, where and for whom.
Add sections with your important pages: services, prices, contact. Next to each link, a short note.
Save it as llms.txt and upload it to the root of your site, next to robots.txt.
Open the address in a browser and make sure you see plain text, not the 404 page.
Check that all the links work and that the prices match the site.
Set a reminder to review it when services or prices change. An out-of-date file does more harm than none.

If you’d like to know whether your site is ready for AI search, send us the address. We’ll tell you what makes sense to fix first, even if that isn’t llms.txt.