Ask Google a question and you get ten links and a map. Ask ChatGPT and you get three names. Sometimes five. And there is no page two.
That one difference should matter to anyone running a business in a town. With ten links, even the seventh result gets a chance. With three names, you either hold a place in the answer or, for that customer in that moment, you do not exist. Not lower down. Not at all.
This article is about what you can do about it. No shortcuts, no tricks, and no promise that there is a button somewhere that puts you in there.
Why this matters right now
Trenčín is a European Capital of Culture in 2026. That means people arriving here for the first time who know nothing about the place. Some of them will open Google. But a growing share will open ChatGPT or Perplexity and type something like "where to get a good lunch in Trenčín" or "places to stay in the centre of Trenčín for a weekend".
The difference from Google is that they do not get a list to browse. They get a recommendation. A finished sentence with a few names and a short explanation of why those ones. Most people do not dig any further and go with it.
Trenčín is just the example where this is most visible this year. The same thing is happening in any district town, and it applies to a hair salon as much as to a garage, a guesthouse or an accounting office.
GEO, AEO, and why this is not another trick
Two acronyms have spread around this topic. GEO stands for Generative Engine Optimization and AEO for Answer Engine Optimization. In practice, both describe the same thing: how to get into the answer the AI generates instead of a list of links.
More important than the acronyms is this: it is not a new discipline sitting next to SEO. It is the same work with a different emphasis. Most of what helps you appear in a ChatGPT answer also helps you in Google, because AI draws on the same web that Google indexes.
And above all, it is not a trick. There is no keyword you drop somewhere and no setting you switch on. The whole thing is about having it stated clearly and consistently across the internet who you are, what you do, where you do it, and whether people have relied on you. Nothing more and nothing less.
If you want it in one sentence: GEO is not about fooling the model. It is about making sure a clear, consistent and verifiable answer about you already exists before anyone asks the question.
Where AI gets local recommendations from
When a model decides who to mention, it assembles a picture from several places at once. The main ones are:
- Your own website. The base source for what you claim about yourself.
- Google Business Profile and map data. Address, category, opening hours, ratings.
- Business directories and registers. Zoznam.sk, Firmy.sk, Azet and the like. Nobody may have looked at them in years, but they still exist and can still be read.
- Reviews and mentions off your website. Local media coverage, partner websites, event pages, blogs.
- Active public social profiles. Not the follower count, but whether the profile is alive and whether the details on it match everywhere else.
What matters is what the model does with all of it. It is not looking for the single best source. It is looking for agreement between sources. When three independent places agree that you do the same thing at the same address, that reads as reliable. When sources contradict each other, the model does not work out which one is right. It simply recommends somebody else who does not come with that problem.
This is the most common reason a business is missing from the answers. Not a weak website. A conflict in the details that nobody inside the company knows about.
Test yourself in ten minutes
Before you change anything, find out where you stand. You need no tool and no budget, just ten minutes and three open windows: ChatGPT, Gemini and Perplexity.
In each one, type the questions the way a customer would. So not your company name, but the problem you solve:
- "Recommend a [your service] in Trenčín."
- "Who does [your service] around Trenčín and are they any good?"
- "I am looking for [your service] in Trenčín, who would you recommend and why?"
- "Compare the options for [your service] in Trenčín."
For each answer, note three things. Who gets mentioned. How they are described, meaning whether the description matches reality. And which sources the model is drawing on, where it names them, which Perplexity always does and ChatGPT often does.
If you are not there, it is not a verdict. It is a starting point. If you are there but described inaccurately, you have a specific thing to fix, and you can usually tell exactly which source that inaccuracy came from.
Repeat the test once a month. Answers change, models get updated, and what you need to see is the trend, not a single snapshot.
The basics: make every source say the same thing
Unify your business identity everywhere. The exact name, address, phone number and company ID. On your website, in Google Business Profile, in directories, on Facebook and on Instagram. The findings are always the same ones in practice: a directory still carries the address from before you moved, somewhere there is still a sole trader name instead of the limited company, and Facebook lists a phone number nobody has answered in years. Every one of those small things reads as a contradiction to the model.
Fill in Google Business Profile completely. The right primary and secondary categories, a description that says what you do and who for, real photographs, and opening hours including public holidays. A half filled profile is the most expensive time saving there is when it comes to local recommendations.
With reviews, what they say matters more than how many there are. Five reviews in which people write specifically about what you helped them with say more than fifty star ratings with no text. So ask happy customers not for a rating, but for a sentence about what you solved for them.
Make your website answer who, what and where in the first paragraph. Not in the About section, not in the footer. In the first paragraph on the homepage. Then: a separate page for each service rather than one shared page with bullet points, and a frequently asked questions section with concrete answers, price ranges and timeframes included.
And one technical point that is easy to miss: the content has to be machine readable. That means the text sits directly in the page HTML rather than being hidden inside an image or added by a script after the page loads. Alongside it come structured data of the LocalBusiness type, an invisible block in the code that states outright that this is a business, with this address, this phone number and these opening hours. This is the part where you usually need somebody who does web development, but it is one off work, not a monthly service.
Beyond the basics: content that can be quoted
Write things that can be picked up. A model quotes specific claims, not the mood of the text. "We provide professional services of the highest standard" cannot be picked up, because it says nothing. "A repair of this kind takes two to three days and costs from 120 euros" can be, and that is exactly the shape of answer the AI is looking for.
What works: concrete numbers, "from" price ranges, a step by step description of the process, and local context, meaning which districts and villages you actually cover. What does not work: generic text that would fit any company in the field. It does not get quoted because there is nothing in it to quote.
Build mentions beyond your own website. What you write about yourself is one source. What others write about you is verification. Local media, the websites of partners and suppliers, pages for events you support, the sites of clients who name you. This is not about buying links. It is about your name existing in places where you are not the one writing it.
How to tell whether it is working
There is no direct report yet telling you how many times ChatGPT recommended you today. You can measure it indirectly, though, as long as you watch several things at once.
In Google Search Console, watch branded searches, meaning people searching for your company name directly. Somebody who learns your name from an AI answer very often goes and checks it in Google afterwards, so a rising number of those searches is one of the best signals available today. Alongside it, watch direct traffic to your website, ask new customers how they found you, and repeat the ten minute test once a month.
None of these is proof on its own. Together, after a few months, they give you a clear picture of whether your visibility is improving.
What does not work
A short list of things not worth spending time or money on:
- Meta keywords. A tag Google has ignored for more than a decade. AI search engines ignore it too.
- Keyword stuffing. Text that repeats "hair salon Trenčín" ten times reads worse and gives the model nothing extra.
- Mass produced articles with no substance. Twenty pieces a month that say nothing specific add no authority. One piece that genuinely answers a customer question does.
- Bought links. A risk for classic SEO, and no gain at all for source credibility.
What they have in common is that they try to get around the system instead of giving it what it is looking for. With a model that compares sources and looks for agreement, that works even worse than it does with Google.
Where to start
If you have an hour for this, do two things: test yourself in the three assistants, and check that your details agree across every source. Those two steps surface most of the problems that can be fixed in an afternoon.
If you want an overview of your online footprint in one place, that is what we built onlinestopa.sk for. It goes through your visibility across the web and returns a concrete plan, not just a list of problems.
And if you would rather not dig through it yourself, get in touch. We will go through how AI and Google actually see you today and tell you what makes the most sense to do first for your particular business. The consultation is free and comes with no strings attached. Sometimes the outcome is that your basics are fine and there are two small things to tidy up. That is useful to know too. If it turns into a bigger plan, we can look at it through strategy.
