If a Slovenian business asks which AI search engine "sees" it, the honest answer is: it depends on the engine. The three most common ones (ChatGPT, Google AI Mode and Perplexity) often answer the same question about the same business differently. In June 2026 we measured this: we asked all three the same questions about 20 Slovenian businesses and counted who recommends whom.
Here is the number that matters: when a customer asks "who do you recommend for this service in this city", Google AI mentions 13 of 20 businesses, ChatGPT only 6.
What we measured
We picked 20 Slovenian businesses from 10 trades customers actually search for (plumbers, air conditioning installers, electricians, accounting firms, car mechanics, hair salons, carpenters, construction firms, dentists, restaurants). Two businesses per trade, from six cities. All winners of classic Google: first page for their trade and city, their own website and a Google Business Profile.
We then asked the three engines the question a real customer types: "who do you recommend in this city for this service". 89 queries in total, all on 12 June 2026, in Slovenian.
Ranking by visibility
This is a ranking on a single criterion: how often the engine mentioned a first-page Google business in a recommendation at all.
| Engine | Businesses mentioned | Share | Note |
|---|---|---|---|
| Google AI Mode | 13 / 20 | 65% | Direct access to Google's data and Business Profiles |
| Perplexity (partial) | 4 / 9 | 44% | Free account cut off after 9 queries |
| ChatGPT (default model) | 6 / 20 | 30% | Leans on business registries, often does not see Google reviews |
Six businesses (30%) were not mentioned by any engine we tested. Not once. Only two businesses out of twenty were visible on all three, both with a high rating on their Google Profile and enough reviews.
Why the answers differ so much
Each engine builds its answer from different sources. We counted the sources across all 89 answers (464 citations in total), and the difference is obvious:
| Engine | Main sources | How it shows in the answer |
|---|---|---|
| ChatGPT | business registries (Bizi, CompanyWall, eBonitete) | Answers read like a credit report. Often does not consider Google reviews. |
| Google AI Mode | Google Business Profiles, company websites | Most citations (248). Recommends freely, shows profiles right in the answer. |
| Perplexity | directories, forums, pages actually visited | Densest citing (12 sources per answer). Twice recommended the directory itself. |
Perplexity has one more quirk worth knowing: on the free account it closed our access after 9 queries for the day. The average user therefore does not get unlimited AI answers, which is itself a data point about how often it can even recommend you.
So which engine is "the best"
For a business owner, that is the wrong question. None of the three is reliable enough to lean on alone. Google AI mentions you most often, but it leans on your Business Profile; if that points to a directory instead of your site, you fall there too. ChatGPT is stricter and reads registries, so it can leave you out or describe you wrongly despite an excellent Google rating. Perplexity may recommend you, but fewer users get far enough to see it.
The practical conclusion: check all three. If one leaves you out, another may recommend you, and only comparing the three shows you where your gaps are.
How to check this yourself
- Open ChatGPT, Google AI Mode and Perplexity and ask each the same question: "who do you recommend for [your service] in [your city]". Write down whether you are mentioned.
- For each, look at which sources it leans on. If they are directories and registries you are not on, you know where to start.
- Repeat after a week or two. Answers change; one measurement is a snapshot, not a rule.
If you want us to measure this for you
We run this exact comparison, focused on your business and your competitors, as a Visibility Teardown: where each of the three engines sees you, which sources drive the answers in your niche, and what to fix first. It costs €290 and is fully credited if you continue working with us. No sales pressure: you can run the steps above yourself.
Limits of the study
Cards on the table: the sample is 20 businesses, enough to show patterns, not statistics. All queries ran on a single day, and AI answers change weekly. Each question was asked once, with no repeats. ChatGPT ran in a temporary chat on the default model, Google AI and Perplexity on logged-in accounts. After 9 queries Perplexity cut off our free access for the day, so its numbers are partial. We do not name the businesses; every example comes from real answers, kept in the study archive.
