The German Medicines Advertising Act and AI marketing: what doctors may and may not do

Why AI visibility suddenly matters for doctors

Your patients no longer just google, they ask. "Which dermatologist in Munich does mole screening with dermoscopy?" or "Is there a cardiologist near me who specializes in heart rhythm disorders?" People type questions like these into ChatGPT or Perplexity today, or they read the AI summary right at the top of Google. The answer names specific practices, focus areas and recommendations. If your practice does not appear there, you simply do not exist for that patient.

Generative Engine Optimization, GEO for short, is the effort to appear in exactly these answers. Unlike classic SEO, it is not about the number one spot in the results list, but about a language model recognizing your practice as a trustworthy, verifiable source and building it into its answer. For doctors this is especially sensitive, because AI providers treat health information as "Your Money or Your Life" topics and scrutinize it more strictly.

That means AI systems are more distrustful when it comes to medical claims. They prefer sources that are verifiable, factual and carry clear authorship. Sensational promises of a cure not only fail legally, the models actively avoid them. This is exactly where the German Medicines Advertising Act meets the ranking preferences of the AI.

What the German Medicines Advertising Act actually prohibits

The HWG regulates advertising for treatments, procedures and medicines. For you as a doctor, three areas are particularly delicate. First, misleading advertising under Section 3: you may not promise an effect that is not established and may not hold out a cure with certainty. "We will cure your migraine permanently" is a clear violation. Second, advertising with fear under Section 11, for example wording that frightens patients in order to push them into treatment.

Third, and this is often underestimated: the ban on advertising with before-and-after images for surgical, plastic-surgical procedures, as well as restrictions on third-party recommendations and expressions of thanks. Expert opinions, testimonials and scientific technical terms that create a false impression of a special effect are also regulated. A dentist may therefore not advertise with a staged before-and-after image of a bleaching treatment if it falls into the realm of a surgical procedure.

Important for your understanding: the HWG does not prohibit information, it prohibits promotional, misleading advertising. Factual explanation of your services, your qualifications and your focus areas is expressly permitted. And it is precisely this factual explanation that is the raw material from which AI answers are built.

The surprising harmony: the HWG and AI want the same thing

Picture the two requirements side by side. The HWG demands: factual, verifiable, no exaggeration, clear limits on the effect. A language model prefers: content with verifiable facts, precise language, a clear source, no marketing waffle. That is practically the same checklist. Where a copywriter once wrote "the best practice in town", you win with the AI using "specialist dermatology practice with a focus on skin cancer screening, digital dermoscopy since 2018".

The reason lies in how the models work. They extract entities and facts: specialty, location, procedure, qualification, opening hours. A statement like "pain-free root canal treatment thanks to modern technology" is worthless to the AI, because it contains nothing verifiable and sounds promotional. "Root canal treatment under the dental microscope with mechanical preparation", on the other hand, is a clean, extractable fact, HWG-compliant and AI-friendly at the same time.

This is the central insight for doctors: you do not have to choose between legal certainty and visibility. The wording your lawyer waves through is usually exactly the one that gets you into the AI answer.

Concrete wording: permitted and prohibited

Let's take real examples from everyday practice. Prohibited and harmful for AI: "Finally live pain-free, our therapy is guaranteed to work." This statement violates the ban on advertising with efficacy guarantees and is ignored by AI systems as an unsubstantiated marketing claim. Permitted and AI-ready: "We treat chronic back pain with multimodal pain therapy, among other approaches. Treatment success varies from person to person and is discussed in the initial consultation."

Second example, aesthetics: prohibited is the before-and-after image of a wrinkle injection with the text "This is how young you'll look afterwards". Permitted and sensible: a factual description of the service such as "Wrinkle treatment with hyaluronic acid, outpatient, consultation on risks and limits included." The AI pulls the procedure, the conditions and your commitment to informed consent from this description, and rates that as a sign of seriousness.

Third example, patient testimonials: gushing testimonials like "Dr. X saved my life, a miracle doctor!" are delicate and come across to the AI more like manipulated marketing. Neutral, verifiable Google reviews with real substance about waiting times, information and friendliness, by contrast, are valuable, legally unproblematic and a strong trust signal for the models.

How to make your practice readable for AI

The technical basis is structured, machine-readable information. On your website, use the schema markup "MedicalClinic" or "Physician" with specialty, address, opening hours and the services offered. This way the AI understands not only that you exist, but exactly what you offer and for whom. A cardiologist who lists their focus areas, such as echocardiography, long-term ECG, heart failure clinic, as separate, clearly named service pages will be cited far more precisely.

On your website, answer real patient questions in plain language. Create an advice section where you factually answer questions like "When should I see a cardiologist about heart palpitations?" or "How does a skin cancer screening work?" This question-and-answer structure is the favorite format of language models, because it can be taken directly into an AI answer. Stay strictly informative, not promotional, that protects you from the HWG and strengthens your findability at the same time.

Pay attention to consistency across all directories: Google Business Profile, jameda, Doctolib, your own website. When name, specialty and address are identical everywhere, the models' trust in your data rises. Contradictory information leads the AI to leave you out rather than risk an error.

E-E-A-T: why medical authority matters for AI

On health topics, both Google and the language models judge the trustworthiness of the source especially strictly. The principle is called E-E-A-T: Experience, Expertise, Authoritativeness, Trust. For you as a doctor this is a home game, if you make it visible. On every professional article, name the author with full name, specialist title and qualification. An anonymous advice page is weighted more weakly by the AI than a text with clear medical authorship.

Link your articles to professional societies, guidelines or reputable sources such as the AWMF guidelines. When your text on skin cancer screening refers to the official recommendations, you signal to the AI that you stand within the scientific consensus. That is legally clean, because you are informing factually, and it raises your authority in the eyes of the models. These very references reduce the likelihood that the AI considers you a dubious source.

Press coverage, talks and teaching also pay in. When you are cited in a professional article or write a guest contribution, mentions of your name appear online. Language models build a picture of your reputation from many such signals. For doctors the rule is: professional visibility outside your own website is often more valuable than any advertisement.

The typical mistakes doctors pay dearly for

The most common mistake is trying to impress the AI with superlatives. "Leading practice", "Germany's number one", "revolutionary method" - such wording is vulnerable under the HWG and is happily challenged by competitors with cease-and-desist letters. At the same time it does not help you with the AI, because the models recognize unsubstantiated superlatives as marketing and filter them out. So you risk a warning letter without gaining any visibility advantage.

A second mistake is outsourcing the text creation to a pure marketing agency without any review under medical advertising law. What sounds good promotionally is often a minefield in the HWG context. Have every text that contains treatment successes, procedures or comparisons reviewed for compliance with medical advertising law. The good news: the reviewed, defused text is almost always the better GEO text too, because it is more factual and richer in facts.

The third mistake is passivity. Many doctors think their good reputation speaks for itself. In the AI world that is no longer true. If you do not provide structured, factual information, the AI fills the gap with data from review portals, forums or outdated directories, over which you have no control.

Your roadmap for the coming weeks

Start with an inventory. Ask ChatGPT, Gemini and Perplexity specifically about your specialty in your location: "Good orthopedist in Regensburg for sports injuries?" See whether and how you appear and what is said about you. These answers are your starting point and often reveal surprising gaps or errors that you should correct.

After that, tidy up your data foundation: consistent details in all directories, clean schema markup on the website, clearly named service pages and a factual advice section with real patient questions. Test every statement against two questions: is it HWG-compliant, meaning factual and verifiable? And is it concrete enough for an AI to build a useful answer from it? If the answer to both is yes, you are on the right track.