How to get named in ChatGPT and Perplexity answers — and why it is not SEO

A customer who used to search for "best car service Prague" now increasingly asks ChatGPT: "who would you send me to for a service in Prague?" The answer arrives as one paragraph containing three names. Either you are among them or, for that person, you do not exist — there is no second page of results here.

Optimising for this is called AEO (Answer Engine Optimization), or GEO. It shares part of its foundation with SEO, but the goal is different: not a position in a list, but a mention in the answer.

How an answer that names you comes about

Simplified, one of two things happens. The model either draws on what it absorbed during training, or it searches the live web while answering and summarises what it finds. Perplexity and the search modes of ChatGPT do the latter — which is why you see links.

The practical conclusion follows: to be named, there must exist content about you that is easy to read, verify and quote. Not marketing copy — facts.

Six things that work

1. Facts on the page instead of adjectives

"We are a modern, dynamic company with an individual approach" is not information. There is nothing in it for a model to extract.

"Car service in Beroun and Prague, specialising in transmissions, open Mon–Fri 8–18, rated 5.0 from 41 reviews, servicing VW and Škoda" — that is five quotable facts. Write pages so that one verifiable statement can be lifted from every paragraph.

2. Answers phrased the way people ask

A question-and-answer section is the single most useful format for AEO: the question as a heading, a short direct answer immediately below it, then the detail. Models look for the question-answer pair, and this structure is the one they take most readily.

3. Structured data

LocalBusiness, FAQPage, Article, Organization and Service markup states outright, in machine-readable form, what a human infers from the layout. It is the cheapest item on this list and the most frequently skipped.

4. Mentions on other people's websites

Models trust repetition across independent sources more than what a company writes about itself. In Czechia that means specifically: Firmy.cz, Mapy.cz, industry directories, professional chambers, local media, podcasts, interviews. One quotation in someone else's article carries more weight than five pages on your own site.

5. Consistent details

If your company name, address and line of business differ from source to source, the model cannot assemble a single entity and will name a competitor whose details are unambiguous instead. Identical NAP everywhere is as fundamental to AEO as it is to local SEO.

6. An llms.txt file

An emerging convention: a text file at the root of the site in which a company states plainly what it is, what it does and where the important pages are. It is not yet a standard anyone is guaranteed to read — but it takes fifteen minutes and has no downside. Ours is at leadai.cz/llms.txt.

How this differs from SEO

The last row is the most important and the hardest. Measuring AEO today means asking the models the same questions manually or by script at regular intervals and recording who gets named. There is no "Search Console for ChatGPT" yet.

How to check yourself

Write down ten questions a customer would ask about your service — in plain language, not keywords. Put them to ChatGPT, Perplexity and Gemini. Record who gets named.

Two things become clear immediately. First, whether you appear at all. Second, and usually more valuable, who the model treats as an authority in your field. That is the list of your real competitors in 2026, and it often differs from the one you expected.

Where to start: a question-and-answer section on the site, structured data, and consistent details across directories. Three steps that improve classic SEO as a side effect — and that, for AEO, decide whether a model can name you at all.