What does ChatGPT say about your business?

Ask two AI models today what they say about your company — free, in thirty seconds. Open the free version of ChatGPT and Gemini, and type exactly what one of your customers would type: "what's the best [service] in [city]?" Most Hungarian SMEs get one of three outcomes here: the model doesn't mention them at all, it recommends someone else instead, or — most often — it simply makes something up. Below I'll walk you through, step by step, how to ask well and how to read what you see.

Before anyone sells you anything, it's worth seeing the situation with your own eyes. You don't need a tool, a subscription, or a specialist for this — just a browser and three minutes. This article is a self-check guide: by the end you'll have a clearer picture of what a customer sees about your company today than you'd get from many expensive reports.

How to ask the right way

The key is not to search for your company name first. No customer searches that way. A customer searches for their problem or for the service, and wants to know who the machine recommends. That's why it's worth trying three different types of question, because each one reveals a different kind of truth.

The first is the blind recommendation question. Here you pretend you've never heard of your company:

"What's the best [service] in [city]? Give me 5 specific names and websites."

This is the most important question, because it's the one a new customer asks. If your name isn't among the five, you're effectively invisible to new prospects.

The second is the reputation question. Here you ask about your company name directly:

"Is [company name] reliable? What's known about them?"

This shows what the model knows — or thinks it knows — about your company. Sometimes it turns out it's confusing you with a similarly named business, or it gives outdated or even false information. This is especially revealing.

The third is the problem question, phrased the way a real customer would when they don't yet know who to turn to:

"[Specific problem] — who should I go to in [city]?"

Where should you ask all this? In three places, because customers are split across them too. ChatGPT's free version is the most widely used. Gemini handles Hungarian best and is closely tied to Google's results, which makes it especially important in the Hungarian market. And Perplexity works with real-time search, showing you which sources it draws from. Ask the same three questions in all three places, and compare the answers.

What does it mean, what you see?

The answer can be roughly one of three kinds, and it's worth recognizing each, because they mean very different things.

Recommended by name. The model says your company's name, gives the correct website, and maybe even adds a sentence about what you're good at. This is the rarest outcome. If you see it, you're among the few — most local businesses never get this far.

Not listed. The model names five companies, but yours isn't among them — or it dodges the question entirely and gives only general advice: "read reviews, request a consultation." This is the typical case. It's not the company's fault, and it doesn't mean your website is bad. It simply means the model doesn't have enough signals to recommend you by name with any confidence.

Says something false. This is the most common outcome, and at the same time the most dangerous. The model confidently provides a name and a website address — one that, when you look it up, doesn't exist, is misspelled, or belongs to another company. The machine didn't know the answer, so it manufactured one. This is called a hallucination. The harm is obvious: the customer would call a phone number that doesn't exist, or end up at your competitor while thinking they're looking for you.

How real is this? I didn't assume — I measured it. I queried four free models about fifteen Budapest dental practices, across three geographic granularities, in 48 queries in total. The result: not a single free model recommended a single real practice in a way where the name was verifiable and reliable — and all four models gave a different list. You can read the detailed case study in zero out of fifteen — AI recommendations in dentistry (my own, dated measurement, May 2026).

What can you do if you're not listed?

Here comes the part where I have to be honest, because it would be easy to turn this into a quick promise. The first three steps, in a nutshell, are the following — and none of them is magic.

1. Make your site readable for AI bots. Permitted crawler access, clean structured data, answer-ready content. This is the entry ticket — without it, the model can't even reach your page.
2. Build your off-site presence. Reviews, independent mentions, appearances in credible sources. This is what decides the recommendation, and it's the slowest to build.
3. Become visible in Hungarian. To win Hungarian customers, compete with Hungarian-language content and mentions tailored to the local audience, not with translated English.

And here comes the most important honest framing, because it's what determines whether someone gets a fair offer or a false one. The technical cleanup grants eligibility — the ability for the AI to find and read your page at all. But that's not what decides the recommendation; off-site presence does: the volume of reviews and appearances in credible sources. International studies show that around 85% of AI citations come from third-party sources, not from the company's own website. Anyone who promises that a good score will make ChatGPT recommend you is selling something that isn't theirs to sell.

Your competitors are visible to AI by accident — their reputation swept them there. The goal is to make your business visible on purpose: by measurement, not by promise.

If you're curious exactly what these seven dimensions measure, from crawler access to off-site presence, and why off-site presence carries the most weight, I lay it out point by point on the methodology page. And you can follow the steps of the process — from measurement to roadmap — on the how it works page.

But the best first step is still to ask the machine right now. Thirty seconds, and you'll see with your own eyes where you stand today. The question is no longer whether AI is paying attention to you, but what it says about you when a customer asks.

And if you'd like a dated, verifiable measurement of your own company — using exactly the method I described above — let me know on the contact page. I'll measure what AI says about you today, and show you where the open ground is.

Kindl Dávid · Published: May 25, 2026 · Updated: June 4, 2026