llms.txt: the file that decides whether ChatGPT mentions you at all

More and more often, customers don't look for you on Google — they ask ChatGPT or Perplexity: “Who does X in Ljubljana?” If the AI can't find a clear answer on your site, it doesn't mention you. The fix is a surprisingly small file — and you can set it up within an hour.

What llms.txt is and why it exists

For decades, websites have had a robots.txt file that tells search engines what they may read. llms.txt is its cousin for the age of language models: a simple text file in the root of your domain (tvojadomena.si/llms.txt) where you write, in your own words, who you are, what you offer, what it costs and what the AI should remember.

The difference is fundamental. Robots.txt is a prohibition. llms.txt is a summary you write yourself — instead of leaving the model to guess from fragmented text scattered across your site.

It's important to be honest about its status: llms.txt is not an official standard that every model is guaranteed to read. It's a proposal that is spreading fast. But the effort-to-benefit ratio is exceptional: an hour's work, no risk, and in the worst case you end up with a tidy summary of your own offer that comes in handy elsewhere too.

Why this matters at all

When an AI answers a question about a service, it doesn't show ten blue links. It gives one answer with two or three sources alongside it. Either you're in that answer or you're not.

In practice, three things decide whether a model mentions you:

• A clearly stated line of business and area. “Digital solutions for modern businesses” tells a model nothing. “I set up AI agents for salons and tradespeople in Slovenia and Croatia” tells it everything.
• Prices, or at least ranges. Questions like “how much does X cost” are among the most common. If you have no number anywhere, you're not in the answer — but the competitor who has one is.
• Answers in a quotable form. Short question, short answer. Three sentences, not three paragraphs.

What it looks like — an example you can copy

The file is plain Markdown. No special tags, no code. A structure that works:

# Company name — what you do, in one sentence

> A two-to-three-sentence summary: who you work for, what you solve,
> in which area and in which languages.

**Contact:** email · website
**Location:** city / service area

## Services and prices
- **Service A — from 690 €, turnaround 5–10 days.** What it includes.
  What it does not include.
- **Service B — from 990 €, turnaround 1–2 weeks.** What it includes.

## How working together goes
1. Introductory call.
2. Written quote with a fixed price and deadline.
3. Build, testing, handover.

## Frequently asked questions (short answers for citing)
**How much does X cost?** From 690 €, turnaround 5–10 days.
**Who is this not suitable for?** For …

## Key pages
- https://domena.si/ — what is on it
- https://domena.si/cenik — all packages with prices

Where to upload it

1. Name the file exactly llms.txt (lower case).
2. Upload it to the root of your site — where index.html or robots.txt lives.
3. Set permissions to 0644 so it is publicly readable.
4. Check in a browser: tvojadomena.si/llms.txt should open as plain text.
5. Don't add anything special to robots.txt — just make sure you're not blocking AI crawlers if you want to be cited.

What else I do alongside it

llms.txt is the first step, not the only one. In an AI-visibility review, there are always three more things besides it:

• Schema markup (JSON-LD) — structured data about the business, services and prices. Both Google and AI systems read it.
• Restructuring the content — I break long paragraphs into questions and answers, because those are far easier to cite.
• Measuring the baseline — I literally ask ChatGPT, Perplexity and Google AI Overviews the questions a customer would ask, and note down who shows up. A month later I repeat it and compare.

Without that last point, there is no way of knowing whether anything helped. The first measurement takes half an hour and is the most instructive part of the whole process — it often turns out the AI is saying something about the business that simply isn't true, because it picked up outdated content from social media or a directory.

Where to start

If you have one hour: write your llms.txt from the template above and upload it. If you have two: add questions and answers, and check what ChatGPT says about you today. These are steps anyone can do themselves, without a developer.