Content & Answer Pages · 9 min read · July 15, 2026
What patients really ask the AI: a search-data analysis for specialists
When a patient today researches a headache, a lump in the neck or a second opinion, they increasingly type it not into Google, but ask ChatGPT or Gemini in full sentences. These questions are a treasure trove: they show you word for word what moves your patients - and whether your practice appears in the answer or not.
Why patients ask differently than they google
At Google a patient types "dermatologist Munich" or "mole dangerous". Short keywords, because they know the machine only understands keywords. At ChatGPT, Gemini or Perplexity something different happens: the patient phrases it like a conversation in the waiting room. "For three weeks I've had a mole on my back that has changed and sometimes itches. Should I be worried and which doctor do I go to?" That's no longer a keyword, that's a fully articulated medical history that tells you more than any click statistic.
For you as a specialist this is a completely new data source. These questions contain symptom combinations, concrete fears, medical histories and the real pressure to act behind them. Whoever reads them systematically often understands their patients better than classic market research ever could. And whoever understands which questions are really being asked can specifically ensure that their own practice gets named in the AI answer as a fitting point of contact, instead of staying invisible.
This is exactly where the difference between classic SEO and Generative Engine Optimization, GEO for short, lies. It's no longer just about position one on Google. It's about whether the AI incorporates your name, your practice or your expertise into its answer at all when a patient in your region poses a health question. In future, visible is whoever gets cited by the machine.
What patients concretely ask the AI
The questions to AI systems follow recognizable patterns. There's symptom interpretation: "I have heart flutters when falling asleep, is that dangerous?" Then specialist assignment: "Which doctor do I go to for constant ringing in the ears?" There's the second-opinion question: "My orthopedist wants to operate on my disc, do I really need that?" And the preparation question: "What should I ask my cardiologist at the first appointment?" Each pattern is a chance to answer with fitting content.
Especially revealing are the fear-driven questions. "Can a PSA value of 4.5 mean cancer?" or "Is a lump in the breast always malignant?" Here the patient seeks not just information, but reassurance and a next step. If your practice website answers exactly these questions factually, empathetically and with a clear recommendation for action, it delivers the AI exactly the material it builds into its answer - including a reference to your specialty.
Organizational questions are also piling up: "Which dermatologist in Regensburg accepts new statutory-insurance patients?" or "Where do I get a short-notice appointment with the neurologist?" These questions are highly local and highly commercial. Whoever is findable here with current, structured details on consultation hours, insurance status and appointment scheduling gets preferentially named by the AI. Precisely such details are completely missing from most doctors' pages.
Where you get this search data
The bad news first: OpenAI and Google don't give you a ready-made list of patient questions. Unlike Google with the Search Console, there's still no official dashboard for AI answers. Nonetheless you're not blind. You can start with real patient dialogues: note over four weeks which questions you and your team really get asked at reception, on the phone and in the consulting room. These are exactly the phrasings that get typed into the AI too.
The second path is direct testing. Put the typical questions of your patients to ChatGPT, Gemini and Perplexity yourself and observe who gets named in the answer. Does a professional association appear, a clinic portal, a colleague, a review portal? Or no one at all from your region? This simple test shows you in black and white where your visibility gap lies and against which sources you're competing.
Additionally, tools like "Answer the Public", Reddit forums, Gutefrage threads or Google's own search suggestions help. There people phrase their health worries as freely as they do to an AI. Collect these real phrasings in a simple table, sorted by symptom, fear and desire to act. This list is the raw material from which your content and your AI visibility later emerge.
From search data to content the AI cites
An AI builds your practice into its answer when you answer the question better than other sources. Concretely that means: for every recurring patient question you need your own, cleanly structured text section on your website. Not an advertising text about your practice, but a real answer. "Are heart flutters dangerous?" is answered with a clear framing, warning signs, self-help and the sentence stating when a cardiologist should be consulted.
The form matters. AI systems love clear structure: a precise question as a heading, a direct answer in the first sentence, then details. Technically correct, but in the patients' language. Avoid pure medical jargon, because the AI matches your phrasing with the user's question. The closer your text is to the real patient question, the more likely it gets drawn on as a source.
And then comes the decisive building block: trust signals. Name the responsible doctor with name and specialist title, link the source of the medical statement, keep the text current. AI models prefer visibly professionally accountable health content. An anonymous blog text has hardly any chance against a page that clearly shows: here writes Dr. med. Weber, specialist in cardiology, as of this year.
The E-E-A-T principle for medical content
Google and the models behind it rate health content especially strictly, because it belongs to the so-called "Your Money or Your Life" topics. Bad medical advice can cause harm, so the systems filter hard for credibility. The keyword is E-E-A-T: experience, expertise, authority and trustworthiness. For you as a specialist that's rather an advantage, because you can credibly deliver exactly these signals, an anonymous advice blog can't.
In practice that means: every medical text needs a named medical author with qualification. A detailed imprint, a real practice address, references to guidelines or professional associations and a visible update date. External signals count too: if you're mentioned on clinic pages, in specialist directories or reputable portals, your authority rises in the machine's eyes. These mentions act like recommendations.
The thinking error of many practices is to hide their own expertise. There it says "our team" instead of a name, no title, no date. To patients that may seem modest, to the AI it's a missing trust signal. Do the opposite: show who writes, what they stand for and where the knowledge comes from. Modesty is a visibility brake in the GEO age.
Think about data protection and professional law
As a doctor you operate within a tighter frame than a normal online shop. The German Medicinal Products Advertising Act limits how you may advertise, and the professional code sets limits on promises of success and misleading statements. For your AI-optimized content this means: inform factually, don't tout. "We cure your tinnitus" is taboo, "This is how tinnitus diagnostics work in our practice" is clean and still gets gladly cited by the AI.
When collecting patient questions, strict restraint applies. Never use real patient data, no names, no identifiable cases in your texts or in AI tools. Work with anonymized, generalized question patterns. If you use AI tools for research, don't enter real people's health data there. Medical confidentiality doesn't end at the keyboard, and a data-protection breach weighs heavier than any visibility gain.
It's smart to show these limits openly as a mark of quality. A note that your content is professionally reviewed and doesn't replace a remote diagnosis creates trust with patients and the AI at once. Models rate responsible, clearly framed health information positively. Honesty about the limits of your statements doesn't make you weaker, but more citable.
A realistic roadmap for your practice
Start small and concrete. Weeks one and two: your team collects every patient question that comes up at reception and on the phone. Week three: you test the twenty most frequent questions yourself with ChatGPT and Gemini and note who gets named. That gives you an honest stocktake of your AI visibility, without spending a cent on agencies. This basis is more than most practices ever have.
Then you prioritize. Choose the ten questions with the greatest pressure to act, usually the fear-driven ones and the specialist-assignment questions. For each of these ten questions you write a short, clear answer section on your website, with a medical author, date and structure. Ten good answers beat fifty superficial texts. Quality and accountability matter more here than volume.
Finally you turn it into a routine. Once a quarter you check again whether you show up in the AI answers, add new questions and update the data. GEO isn't a one-time project, but an observing of how patient questions and models change. Practices that establish this rhythm now secure a head start that late-starting colleagues can hardly catch up on.
Conclusion: listening is the new search engine optimization
The core message is simple: patient questions to AI systems are the most honest market research you've ever had. They show you in plain text which worries, symptoms and decisions preoccupy your target group. Whoever collects these questions and takes them seriously understands their patients more deeply and can align their communication precisely with them. That benefits visibility and quality of care alike.
Visibility in the AI arises not through tricks, but through real, professionally accountable answers to real questions. Your task is to prepare this knowledge, which is already in your head anyway, so the machine can find, understand and cite it. Structure, name, date and source are the bridge between your expertise and the AI answer reaching the patient.
Start today with the simplest step: ask your team which three questions came up most often this week. These three questions are your starting capital. Answer them better than everyone else, and you lay the foundation for the next patient who asks their AI for advice to hear the name of your practice.
Common questions
Does AI-optimized content violate the ban on medical advertising?
No, as long as you inform factually instead of touting. The Medicinal Products Advertising Act and the professional code prohibit misleading statements and promises of cure, but not comprehensible patient education. Phrase things neutrally, do without superlatives and guarantees of success and point out that content doesn't replace medical advice. That way you stay legally safe and still get gladly cited by the AI.
How do I know whether my practice shows up in ChatGPT or Gemini?
Test it yourself. Put to the AI systems the typical questions of your patients, such as for a specialist in your field in your city, and see who gets named in the answer. Repeat this with several phrasings. If you're not mentioned, you see directly which questions and sources you're still missing. You should repeat this check roughly once per quarter.
Is the effort even worth it for a small specialist practice?
For small practices in particular, yes. Large clinic corporations move slowly, whereas you can quickly publish ten precise answers to real patient questions. Because AI systems prefer professionally accountable, local and current content, you as a named specialist have a structural advantage over anonymous portals. An early start secures you a head start that can hardly be caught up later.
Read on