Getting found in ChatGPT: GEO for Swiss businesses

Getting found in ChatGPT means: your company has to appear so clearly and consistently across the web that an AI recognizes it as a source, reproduces it correctly and names it in its answer. That is exactly what GEO is, Generative Engine Optimization. I’ll show here what that looks like in practice for a Swiss SME.

The background: more and more people no longer type their question into Google, but ask ChatGPT, Perplexity, or get a summarized AI answer right at the top of Google search (the so-called AI Overviews, visible in Switzerland since spring 2025). Instead of a list of ten blue links, there’s a finished paragraph that already gives the answer. Off to the side: a few sources the AI pulled it together from.

For an SME, that shifts the decisive question. It used to be: am I on page one of Google? Now there’s a new one: do I even get named in this AI answer, and if so, with the right details? I’ll explain here, plainly, what’s behind it and what you concretely need for it. No hype, no firm promises.

What changes compared to classic search

In a normal Google search, the person clicks a result themselves and reads your page. With an AI answer, the machine reads your page first, summarizes it and hands the person the result. The click on your website often falls away, or it only comes once the AI names you as a source and the user wants to know who that actually is.

That has two practical consequences. First: the AI has to be able to understand your content at all and reproduce it correctly. Second: it should cite you as a trustworthy source, not a review platform or a competitor who happens to be better structured.

Much of this still isn’t measurable in any reliable way. There’s no “position 1” in ChatGPT that you can read off every day, and I don’t promise anyone firm numbers. But the direction is clear, and the approach for becoming citable is surprisingly down to earth.

Why an AI gets your company wrong

When an AI today answers the question “Who does carpentry work in the Rhine Valley?”, it pulls the answer from what it finds about you on the web. If that’s patchy, contradictory or outdated, then the answer is patchy too. Three typical problems I keep seeing with Swiss SMEs:

• The details don’t match up. Your own website has the new address, the old business directory the old one, Google the phone number from two years ago. The AI then doesn’t know which is right, and when in doubt names none of them or the wrong one.
• The answer isn’t stated anywhere in a sentence. The information is there, but it’s scattered across a long block of text or buried in a PDF. An AI most likes to pull out a clear, self-contained sentence. If it finds none, it guesses or leaves you out.
• Nobody else confirms it. An AI weighs what shows up consistently in several places. If only your own page makes a claim and no other source backs it, the trust is lower.

What a Swiss SME concretely needs

The good thing about it: what makes a page readable for the AI also makes it better for people. It’s not a trick, it’s ordinary groundwork. Four things matter most.

Clear answer sentences. Your customers’ most important questions should be answered somewhere on your page in a single, unambiguous sentence. Not “we offer a broad spectrum of solutions”, but “We repair heating systems in the St. Gallen area and get back to you within one business day”. A sentence like that can be quoted, a hollow phrase cannot.

A real FAQ structure. A question as a heading, the answer directly below in two or three sentences. That’s exactly the bite-sized format an AI likes to take over. It also helps the customer who wants to quickly look something up without calling.

Key facts that agree with each other. Company name, address, phone number and opening hours have to be the same everywhere: on the website, in the legal notice, in the Google Business Profile and in the directories you’re listed in. Sounds banal, but it’s the most common source of error. With the legal notice, care is required in Switzerland anyway since the revised Federal Act on Data Protection (revFADP, in Switzerland known as revDSG). A complete, correct legal notice is at the same time a reliable source an AI can draw your master data from.

Machine-readable markup in the background. In the source code, the most important details can be marked up so that a machine assigns them without doubt: this is the company, this the address, this a review, this an FAQ. The visitor sees none of it, but it makes it easier for the AI to reproduce you correctly. That’s the technical part I take care of.

When it’s worthwhile for you, and when not

I’ll say it openly: not every business needs to invest in GEO today. If your customers find you through referral and you’re at capacity, this isn’t a pressing topic. And GEO replaces neither a good website nor the classic search engine work, but builds on it. A page that isn’t found on Google and has unclear content won’t suddenly be recommended by an AI either.

It becomes worthwhile once your customers start asking an AI questions like these, and you want to make sure the answer is right and that you appear in it. Today that mainly affects industries that get researched online before someone calls. The first step usually isn’t expensive at all: check what ChatGPT and Perplexity say about your company today, and close the obvious gaps and contradictions.

Where you can read on

On the page about GEO and AI search I show with examples what such an AI answer concretely looks like, how a single answer emerges from several sources and what makes a page AI-readable. That’s not a sales promise, but a plain overview of something that’s still developing. And when the effort is not worth it is set out separately in When GEO is not worth it for a Swiss business.

Here’s what this looks like from the other side. Somebody types your trade and your region into ChatGPT and gets three names back, and either yours is missing or it appears with the phone number you gave up in 2022. Nobody tells you. There’s no ranking, no report, no email. That’s the uncomfortable part of the topic, and the reason to do the cheap thing first: type your own company name into ChatGPT and into Perplexity and read what comes back. Whatever is wrong in that answer is wrong somewhere out on the web as well, and where nothing comes back at all, the sentence somebody could have quoted is missing. That’s where the work actually starts.