How AI engines choose which medical practice to recommend

Patients no longer just type a symptom into Google. They ask an assistant to pick a doctor for them. Ask ChatGPT, Perplexity, and Gemini the same thing, who is the best concierge doctor in Coral Gables, and you will often get three different shortlists, because the overlap in the sources these engines trust is surprisingly small.

Why do the engines recommend different practices?

Because each engine has its own source pool and its own logic for ranking it. A page ChatGPT trusts may never surface in Perplexity, and a practice that dominates Google AI Overviews may be absent from Claude. They are not the same product with a different logo. Treat each as its own channel and the differences become a map instead of a mystery. The good news is that the foundation is shared, so you are not starting five times over. You build one credentialed presence, then tune a few signals per engine.

The rule that comes first: keep it clean and credentialed

Health content carries rules that other verticals do not. Never publish protected health information, and never name or describe a real patient in public copy without written consent. If you share an outcome, frame it as one person's experience with a plain "individual results may vary" note, and make sure any health claim can be backed up if the FTC ever asks. None of that holds you back. Engines cite specific, verifiable facts anyway: the named physician, the board certifications, the defined list of services. Vague copy that never says who the doctor is or what they treat gets skipped in every engine at once.

ChatGPT: authority and a recognizable physician

ChatGPT leans on authoritative sources, high-ranking pages, patient-review sites like Healthgrades and Zocdoc, and Google reviews, and it cites with inline links beneath the answer. For a prompt like top dentist near Brickell, it rewards a practice it can recognize as a distinct entity with real credentials. The lever is a named physician with board certifications marked up in Person or Physician schema, complete practice schema, and one clear, quotable answer per page. If the model cannot tell who the doctor is or which practice a page belongs to, it moves to a source it can.

Perplexity: recency and a strong opening

Perplexity favors fresh content and recent publications, pulls from community sources, and cites with numbered footnotes. For a prompt like direct primary care physician in Miami Beach, it rewards recency signals, a strong headline, and the key takeaway in the first paragraph. The lever is visible publish and updated dates, an answer-first opening line, and a footprint that looks current rather than last touched two years ago. A practice with excellent but undated services pages often underperforms here for no reason other than the missing date.

Claude: depth and sourcing

Claude prefers long-form, well-structured explainers and surfaces sources inline and at the end of an answer when browsing. It rewards depth over surface and claims that are backed by real sourcing. The lever is thorough pages that explain a condition or a procedure clearly and cite their own sources, not thin FAQ stubs. For a medical practice this plays to a natural strength, because the honest answer to a patient question, what does concierge care actually include, or how does a dental implant timeline work, usually is nuanced. A page that treats it that way reads as the credible source.

Gemini and Google AI Overviews: the Knowledge Graph and local SEO

The Google surfaces reward what Google has always rewarded, plus structured data. Gemini weights the Google index, Knowledge Graph entities, and YouTube. Google AI Overviews pulls top-ranking organic pages and schema-rich content into source chips. For a prompt like concierge cardiologist Key Biscayne, the lever is solid local SEO, a claimed and consistent Google Business Profile, and complete MedicalBusiness, Physician, and FAQPage schema so the practice reads as a real, located entity. Practices that never did the classic local SEO groundwork tend to be weakest exactly here.

The shared foundation under all five

Read down the last column and the pattern is clear. A named physician, complete schema, and a real answer help everywhere. After that you are tuning: recency for Perplexity, depth for Claude, Knowledge-Graph and Google Business identity for the Google engines. None of it is exotic. It is the same credentialed foundation, made legible to five different readers, and most practices are strong in one or two of these and quietly invisible in the rest.

Where to start

Start where your patients already ask. If they live in ChatGPT, the fastest wins are physician credentials in schema, complete practice markup, and a clean answer block on your core service pages. If they compare options in Perplexity, dates and answer-first openings matter more. Then widen out. The work compounds, because the foundation you build for one engine is most of the work for the next, and a practice that is legible to all five, found, trusted, and cited by name, is far harder for a competitor to displace than one that got lucky in a single tool.