How Doctors Show Up in AI Search

Patients have started asking AI assistants for doctors. “Find me a cardiologist near Pasadena who speaks Spanish” is a question ChatGPT, Claude, Gemini and Perplexity all answer today — with names. Which names come back is not chance, and it is not advertising: assistants recommend providers whose information they can actually read, verify and cite. This page explains how that works and what a physician can do about it, in plain terms.

How do AI assistants choose which doctors to recommend?

An assistant answering “find me a knee surgeon in Denver” pulls from what it has seen: structured directories, official registries, hospital sites and pages it can crawl. It prefers sources where the facts are machine-readable — name, specialty, location, languages, services, credentials marked up in schema.org structured data — and where the same facts agree across sources. A provider whose information exists only as a PDF, a photo of a business card, or an out-of-date listing effectively does not exist to the model.

Why doesn’t ChatGPT recommend me?

Almost always one of three reasons. It cannot read you — your information is not on any page with structured data an AI can parse. It cannot trust you — your NPI record, your hospital page and your directory listings disagree with each other, so the model hedges and names somebody else. Or it cannot find enough — a bare name-and-address listing loses to a colleague whose profile answers the patient’s actual question: conditions treated, languages spoken, insurance, philosophy of care.

What is “AI SEO” for doctors — and do I need to pay for it?

You will see it sold as AEO (answer engine optimisation), GEO (generative engine optimisation) or AI SEO, sometimes for four figures a month. Strip the acronyms and it is three chores: put accurate, structured, machine-readable information about yourself on pages assistants crawl; keep it consistent with the federal NPI registry; and make it complete enough to answer real patient questions. None of that requires an agency. Most of it is claiming and completing profiles you already have.

How do I check what AI says about me right now?

1. Ask three assistants the question your patients would ask: “find me a [your specialty] in [your city]” — and then ask each one “what can you tell me about Dr. [your name]?”
2. Note what is wrong, missing, or stale — an old address and a departed practice are the commonest failures.
3. Search your own name plus “NPI”. The federal record is what most directories, including this one, are built from — if it is wrong, everything downstream is wrong. Update it at nppes.cms.hhs.gov.

So what should I actually do?

Claim your profile on directories that publish structured data — like this one. A claimed guide.md profile is identity-verified, marked up with Physician schema an assistant can parse, kept consistent with your NPI record (you approve every registry change before it touches your page), and lets you add the things no registry holds: languages, conditions and procedures, hospital affiliations, publications, a photograph, your own words. That combination — verifiable, structured, complete — is precisely what assistants are built to prefer.

Then do the same anywhere else patients find you: your hospital’s physician finder, your practice site, the big review sites. Consistency across sources is itself a ranking signal.

Will this fix things overnight?

No, and be wary of anyone who promises otherwise. Assistants refresh their view of the web on their own schedules — some retrieve live, some retrain periodically. What you control is whether, whenever they next look, they find a complete, consistent, machine-readable record of who you are and what you do. That is a one-time hour of work plus an occasional glance, not a subscription.