AI Engines · 9 min read · July 15, 2026
How to get your medical practice into the recommendations of ChatGPT and Gemini
More and more people no longer ask Google, but ChatGPT or Gemini: "Which family practice near me is still taking new patients?" If your practice doesn't appear there, you simply don't exist for these patients. Generative Engine Optimization ensures that AI systems know your practice, describe it correctly and actively recommend it, in a serious, fact-based way and within the bounds of the German medicinal advertising law.
Why patients today ask ChatGPT instead of Google
Your patients' search behavior has shifted, more quietly than many practice owners think. Whoever used to type "dermatologist appointment Munich Schwabing" into Google today phrases whole sentences in ChatGPT: "I've had an itchy rash on my arm for weeks, which specialist in Munich should I see and what should I watch for when choosing an appointment?" The AI answers not with ten blue links, but with a concrete recommendation. Either your practice is in it or it isn't.
That is the decisive difference from the classic search engine. With Google you compete for spot three or four and still have a chance at the click. With ChatGPT and Gemini there are often only two or three named names. Visibility here has become binary: named or invisible. Especially in overcrowded specialties like dermatology, orthopedics or psychotherapy, this decides whether new self-pay and private patients even learn of your existence.
A realistic view is important: AI systems don't replace the doctor's visit, but they increasingly become the first point of contact. They pre-filter, they sort, they recommend. Whoever ignores this pre-filter loses patients before the first phone call even takes place.
How ChatGPT and Gemini even get their practice info
Language models don't invent their recommendations, they draw them from sources they consider trustworthy. For medical practices these are above all: your Google business profile, review portals like Jameda or the doctor-review portal of the health insurers, specialist directories like the doctor search of the Associations of Statutory Health Insurance Physicians, your own website, and in part regional press articles. What is stated consistently in these sources lands in the AI's answer.
The problem for many practices is inconsistency. The website says "Practice Dr. Meier & Colleagues", on Jameda "Family practice Meier", in the Google profile an old address from the time before the move. For a human this is forgivable, for a language model it's a reason for uncertainty. Contradictory information leads the AI not to name your practice at all, because it doesn't trust the facts.
Your first task is therefore not marketing, but data hygiene. Name, address, phone number, specialty, opening hours and service range must be identical across all sources. This foundation sounds banal, but it's the lever with the greatest effect.
Which questions your patients really ask the AI
To land in recommendations, you have to know what's being asked. With doctors the queries rarely run through the pure specialty label. Typical are phrasings like "Which family doctor in Cologne-Ehrenfeld is still taking new statutorily insured patients?", "I need a short-notice appointment with a gynecologist in Leipzig, who has free capacity?" or "Which practice in Stuttgart offers acupuncture for migraines?"
It's striking how strongly these questions hang on concrete needs: statutory-insurance approval, accessibility, languages of the practice team, special procedures, short-notice appointments, experience with children or anxious patients. It is exactly these attributes you should name explicitly on your website. An AI can only recommend what exists somewhere as text. If it's nowhere stated that you offer diabetes education or Turkish-language consultation, you won't appear for these questions.
Collect the real questions from your everyday practice. What do callers ask before making an appointment? What's in the emails of prospects? These sentences are the most valuable raw material for your content, because they correspond exactly to the language in which patients also talk to the AI.
Build your website so the AI understands it
Language models like clarity. Long, nested welcome texts full of clichés about "holistic care with heart" don't help them. What helps are unambiguous, fact-rich pages: a service page per specialty, clear headings, running text that answers a concrete question. Instead of "We offer state-of-the-art diagnostics" you'd better write which examination you perform for which complaints and whom it's suitable for.
Real FAQ sections are especially effective. If you answer the question "Do you take new statutorily insured patients?" in a fully worded way on your page, you deliver a ready, citable answer to the AI. Add structured data in the background, the so-called schema markup for medical organizations. With it the machine recognizes without doubt that this is a medical practice with certain specialties and opening hours.
With every page, think of the combination of location and occasion. "Dermatologist mole screening Dresden Neustadt" is more concrete and thus more findable than a general dermatology page. The more precisely your content docks onto real concerns, the more likely you become the fitting answer.
Reviews and reputation as a trust signal
AI systems weigh whom to recommend, and reputation plays a big role in that. A practice with many current, differentiated reviews seems more trustworthy to the model than one without traces on the web. That doesn't mean you need bought floods of five stars, quite the opposite. Conspicuous patterns are recognized both by portals and by models, and in the health sector that harms more than it helps.
Instead, rely on an honest, systematic process. At the end of the appointment, kindly ask satisfied patients for a review, don't idealize, and respond professionally to critical voices. It is precisely the matter-of-fact handling of criticism that is a signal radiating quality. Observe confidentiality in the process: in public replies you may never confirm that someone was even a patient of yours or name treatment details.
Consistency over time beats individual spikes. A steady stream of real feedback over months builds a stable picture that is reflected in AI answers. Reputation here is a marathon, not a sprint, and that is exactly what makes it forgery-proof.
The legal limits: take the medicinal advertising law seriously
Unlike a restaurant, as a doctor you are subject to strict advertising rules. The German Medicinal Advertising Act and medical professional law prohibit misleading, promotional or comparative advertising. Statements like "the best cardiologist in town" or promises of cure are taboo, even if they were in a text optimized for the AI. GEO for doctors means informing factually and verifiably, not exaggerating.
That's not a restriction, but actually an advantage. Language models prefer sober, verifiable facts anyway and distrust superlatives. If you describe cleanly which services you offer, which qualifications your team has and which patient groups you're suitable for, you serve both the AI and professional law at once. Facts beat advertising language in both worlds.
Caution is called for with before-and-after depictions, testimonials about individual treatments and concrete success figures, especially in aesthetic fields. When in doubt, it's worth looking into the guidelines of your State Medical Association or a brief legal check before content goes online.
How to check whether the AI already knows you
Before you optimize, get an honest baseline. Ask ChatGPT, Gemini and also the AI overviews in Google search the questions your patients would ask. "Recommend me a family practice in my district", "Which orthopedist in my town treats sports injuries?" Note whether you're named, whether the information is correct and who is recommended instead.
You'll often experience surprises: outdated opening hours, a wrong specialty or the naming of a practice that no longer exists. Each of these deviations is a concrete task. Also check from which sources the AI draws its information; many systems name them on request. That way you recognize which portal you should update next.
Repeat this check regularly, for example quarterly. AI models are retrained and their answers change. What's correct today may be outdated in six months. This routine costs little time and shows you in black and white whether your measures are working.
A realistic roadmap for your practice
Start with the basics, not the finishing touches. First clean up your Google business profile, reconcile all directory entries and ensure uniform basic data. Experience shows this first step brings the biggest jump, because you thereby eliminate the most common cause of invisibility: contradictory information the AI doesn't trust.
In the second step you overhaul your website along real patient questions: clear service pages, an honest FAQ section, structured data in the background. In parallel you establish a calm, permanent review process. Plan this not as a one-off campaign, but as a fixed habit in everyday practice, which the one person on the team responsible looks after.
Be patient with the results. GEO doesn't work overnight, because models need time to absorb new information. But the practices that start now secure a head start that latecomers will later find hard to catch up on. Whoever ensures today that ChatGPT and Gemini know their practice correctly will be recommended tomorrow, while others are still wondering whether all this is even relevant.
Name specialties clearly
The AI recommends you when it can classify your practice unambiguously. So don't just write "family doctor", but name concrete specialties: diabetology, travel medicine, outpatient minor surgery or disease management programs. The more precisely your website and your directory entries word these services, the more reliably ChatGPT assigns you to a particular patient question.
Think in real patient phrasings. Someone rarely types "diabetologist", but asks "Where in Rosenheim can I get my blood sugar set right long-term?". If your text takes up such everyday language and links it with your technical term, you build exactly the bridge the language model needs.
Avoid scattering your specialties across ten subpages. A clearly structured service overview with short descriptions per offering is easier for the AI to capture than long running texts in which the actual core competence gets lost.
Make availability and admission of new patients transparent
One reason the AI doesn't name a practice is banal: it's unclear whether you even take new patients. Write explicitly on your website whether your patient roster is open, whether there are waiting times and for which insurers you're approved. The model takes this information as a signal to actively recommend you.
Consultation hours, appointment booking and emergency notes also belong clearly worded on a fixed page. If a patient asks "Which practice near me is still open today?", the AI can only answer if your opening hours are cleanly and currently stored, ideally identical in the Google business profile and on the website.
Keep this information consistently maintained. Nothing harms your recommendation more than contradictory information between portals. An outdated holiday note or a wrong phone number acts on the model like an uncertainty signal and pushes you out of the answer.
Answer common patient questions as a mini-roadmap
Set up a short section on your website with real patient questions and answer them directly: How does the first examination proceed? Do I need a referral? How do I get a prescription without an appointment? Such question-answer blocks are ideal for ChatGPT and Gemini, because they correspond exactly to the format in which the AI answers patients.
Phrase every answer in two to three sentences, factually and without promises of cure. Stick to verifiable statements about processes, equipment and organization. That way you stay within the legal limits and still deliver the model concrete, citable substance.
As a small roadmap: collect the ten most common questions from your reception over four weeks, write a short answer per question, publish them bundled and check after a few weeks whether ChatGPT names your practice for exactly these questions.
Common questions
Is it even legally permitted to optimize my medical practice for AI recommendations?
Yes, as long as you adhere to the Medicinal Advertising Act and your professional law. Factual, verifiable information about services, qualifications and opening hours is permitted and even desired. What remains prohibited is promotional advertising, superlatives like "best doctor", promises of cure and misleading statements. GEO for doctors is pure fact maintenance, not advertising, and that is exactly what the AI systems prefer anyway.
How long does it take for ChatGPT or Gemini to recommend my practice?
Don't count in days, but in months. Directory and Google-profile corrections take effect relatively quickly, because many AI answers access these sources live. The models' trained base knowledge updates more slowly. A realistic horizon is three to six months, during which you build up consistent data and real reviews. Check the progress quarterly with your own test questions.
Do I have to spend money on reviews or ads to get named?
No. Bought reviews are risky in the health sector and are recognized by portals and models alike. The most effective lever is free: clean, uniform basic data across all sources and a fact-rich website. An honest review process with real patient voices costs you above all attention in everyday practice, no advertising budget. Paid ads don't influence the editorial AI recommendations.
Read on