A growing share of patients no longer look for their dentist, eye clinic or cardiologist on Google — they ask ChatGPT and Gemini. We ran an experiment: thirteen typical patient questions put to both systems, across eight healthcare verticals. The results matter to every private healthcare provider — and several of them surprised us.
About the research. Conducted in August 2026 on the Hungarian private healthcare market, in Hungarian. Thirteen patient-intent questions across eight verticals — dentistry, orthodontics, ophthalmology, cardiology, gynaecology, paediatrics, psychology and orthopaedics — asked of both ChatGPT and Google Gemini, for twenty-six tests in total. Questions were phrased the way a patient would phrase them, including the city.
Why a non-English market is worth studying. Almost all published research on AI visibility measures English-language markets, where the models have the most training data and the competition for AI recommendations is already fierce. A smaller, non-English market shows what these systems do when the data is thinner and almost nobody has optimised for them yet — which is the situation most industries in most countries are actually in. The names below are Hungarian, but the mechanics are not.
AI names individual doctors — even in a YMYL field
Healthcare is a so-called YMYL area (Your Money or Your Life), where the search giants have traditionally been cautious about recommendations. Against that backdrop, the first finding of our test was this: AI answers recommend boldly, and by name. For the orthodontics question, ChatGPT returned a specific specialist in first place (Dr. Levente Szegedi); for plastic surgery it listed two doctors by name; and for invisible orthodontics Gemini recommended Dr. László Balla, highlighting his twenty years of international experience.
So the patient today receives not only an institutional recommendation from the AI, but the name of an individual doctor — which raises the importance of the personal professional brand in healthcare higher than ever before.
The largest chains are the ones missing
You would assume the biggest players in private healthcare dominate AI recommendations. The reality: Medicover appeared in one answer out of twenty-six tests, and TritonLife in none. Meanwhile a whole series of small private practices with five-star Google ratings made the lists: Gondosorvos and 27 Sellő in cardiology, Nődokik in gynaecology, the Merengő psychology practice — all providers that are a fraction of the big chains in size, but beat them on reviews.
The same pattern emerges here as in our earlier twelve-industry study: neither size nor marketing budget decides the outcome. Digitally documented evidence does — reviews, specialisation, and presence in trustworthy sources.
The other side of that coin is merciless: the AI shows the weak score too. On the paediatrics question, one well-known institution appeared in ChatGPT’s map-based recommendation with 3.8 stars — directly next to competitors at 4.4 and 4.6. On the cardiology question, a long-established health centre’s 4.0 rating stood out among five-star small practices. The patient therefore sees a comparison at the very first point of contact, before reaching any provider’s website. Review management has stopped being a cosmetic question: it has become a direct driver of patient volume.
Set markets and fluid markets
The most interesting pattern appeared when we overlaid the two AIs’ answers on each other. Consensus turns out to differ dramatically by vertical.
For laser vision correction, ChatGPT’s and Gemini’s lists were practically identical — Sasszemklinika, Focus Medical, Budai Szemészeti Központ, Saint James — and for joint replacement surgery both systems put Emineo private hospital first. These are set markets: the players’ positions are stable, but getting in alongside them is very hard.
In cardiology, gynaecology and psychology, by contrast, the two AIs recommended almost entirely different names. These are the fluid markets, where no consensus has formed yet, and where a provider building deliberately can still win a defining AI position relatively quickly. For dental implants, out of the eight to nine named clinics returned by the two systems, exactly one was shared. Whoever moves now is still paying the cheap entry price in the fluid markets.
The clearest positive example: narrow focus wins
The best positive case in the study came from the shockwave therapy question. Asked “I was recommended shockwave therapy for heel pain, where can I get this treatment in Budapest?”, ChatGPT’s map recommendation surfaced Harmónia Centrum — and Gemini went further, recommending the downtown practice in first place, by name and address (“Harmónia Centrum – Shockwave Therapy Centre”).
The explanation is textbook: the practice has positioned itself consistently around that one specialty for years, with several thousand documented treatments and strong patient reviews — and that focused, evidence-backed presence is exactly what an AI learns from. The same pattern recurred with MelanomaMobil (mole screening) and KardioKözpont (cardiology): a narrow specialty focus produces AI visibility even for a small player, while generalists get lost in the field.
Which sources the AI actually reads
The source attributions in the answers are revealing too. Four source types came back systematically: specialised review and directory sites; doctor booking platforms (a significant part of the paediatric and gynaecology recommendations was built from these); the clinics’ own well-structured professional content; and community sources, with Reddit leading the way.
The lesson is unambiguous: if you are not on the booking platforms with an up-to-date profile, and if no independent third-party content is written about you, you barely exist for the AI — however excellent your actual care may be.
There were also questions where the AI visibly struggled to find candidates: for mole screening, ChatGPT could only recommend two names. These ownerless topics are the fastest GEO positions to occupy — where one or two strong pieces of content and a consistent presence can still earn the top spot today. Where such gaps sit differs by vertical, but on the evidence of our test there are several of them.
Three conclusions
First: AI visibility in healthcare is already a real patient-volume question. Patients ask the AI even for their most sensitive, highest-trust decisions, and the AI answers — by name.
Second: positions are not distributed by company size. Well-documented, specialised providers with excellent reviews are ahead of the large chains.
Third: the window will not stay open forever. The verticals that are setting show what happens once a field settles. This is not a one-off configuration but a matter of continuously maintained presence: reviews, platform profiles, professional content and independent sources together, optimised for two AI systems with different logic.
The research raises uncomfortable questions as well. The AI recommends a healthcare provider — but who is responsible if it gives the patient an outdated address, a closed practice or a wrongly attributed service? In our tests the overwhelming majority of answers proved accurate, but the systems themselves warn that they can be wrong. In healthcare the stakes are higher than in any other industry — which is at once an argument for caution and an argument that providers should make sure the AI learns about them from accurate, current information they control.
Written by Endre Kovács, SEO and GEO specialist, managing director of Marketing Kalkulátor Kft. If you would like the same test run on your own market, get in touch.