AI search for medical practices: what changed and what to do about it

For about twenty years, healthcare marketing had a stable shape. Patients searched, Google returned ten links, and practices competed for position. Everyone understood the game even if they played it badly.

That shape is coming apart, and healthcare is among the categories where it’s coming apart fastest — because health questions are exactly the kind of thing people prefer to ask conversationally, in private, without clicking through five pages of results.

Here is a plain account of what changed and what a practice should actually do.

What changed

The interface changed from a list to an answer. When a patient asks an assistant for a recommendation, they get two or three names with reasoning, not ten links to evaluate. The shortlist is formed before the patient ever visits a website. Being on page one of Google is meaningless if the shortlist was assembled somewhere else.

Queries got longer and more specific. Nobody types “dentist near me” into a chat window. They type “I chipped a molar and I’m scared of dentists, is there somewhere in Portland that does sedation and takes Delta Dental.” The practices that get named are the ones whose information addresses each of those constraints.

Google itself changed. AI Overviews now answer above the traditional results for a large share of health-related searches. Even patients who never open ChatGPT are getting AI-assembled answers, which means the shift reaches practices whose patients aren’t early adopters of anything.

Research moved earlier and got longer. Patients researching an elective procedure now have extended, multi-turn conversations weeks or months before contacting anyone. They arrive at the consult with formed opinions about technique, recovery, and cost.

What didn’t change

Quite a lot, which is the reassuring part.

The fundamentals still overlap heavily with good SEO: accurate structured information, a technically healthy site, consistent listings, real reviews, and content that answers questions. The work that makes you citable by an assistant is largely the work that makes you rank in search.

Local intent still dominates. Trust still decides. Word of mouth and physician referral still matter enormously.

What changed is the interface and, critically, the penalty for being unstructured. A practice with disorganized information used to rank poorly. Now it can be omitted entirely from an answer, which is a categorically different outcome.

Why independent practices are mostly invisible

Ask an assistant for the best clinic in a mid-sized metro and you will usually get some combination of hospital systems, large groups, and directories.

This isn’t because those organizations did clever AI optimization. It’s because they have complete, consistent, well-linked information at scale, and independent practices frequently don’t. A model recommending a hospital system is making a safe bet with well-corroborated data. A model considering an independent clinic with unclear hours, an image-based insurance list, and a booking widget it can’t read is making an uncertain one — so it doesn’t.

The gap is a data-quality gap, not a budget gap. That’s why it’s addressable.

The concentration problem

There’s a structural consequence worth understanding: AI answers concentrate attention far more aggressively than search results do.

Ten blue links distributed attention across ten businesses. An assistant naming three concentrates it on three. For the practices named, this is excellent — AI-referred visitors arrive later in the decision, having already been vetted. Adobe Analytics, measuring US retail, found that in July 2026 these visitors converted 60% better than other traffic, spent 59% more time on site, and were 34% less likely to bounce. Healthcare has no equivalent published dataset yet, so treat the figure as directional — but the behaviour it describes is exactly what patient research looks like.

For everyone else, the traffic doesn’t arrive at all. There’s no long tail of position-seven clicks to fall back on.

This makes visibility work higher-stakes than it used to be, in both directions.

What to actually do

Audit what assistants say about you now. Ask each major assistant the ten questions that would fill your schedule. Not your practice name — the questions a patient who doesn’t know you yet would ask. Record who gets named. This takes an afternoon and is usually clarifying in an uncomfortable way.

Fix consistency first. Name, address, phone, and hours identical everywhere. This is unglamorous and it’s the highest-return item on the list, because inconsistency undermines a model’s confidence in everything else it knows about you.

Make your facts machine-readable. Insurance plans as text, not an image. Hours as structured data, not a graphic. Services named individually. Providers with verifiable credentials. Schema markup so nothing has to be inferred.

Answer the questions patients actually ask, completely, in place. Especially cost. The practices that publish honest pricing ranges get cited disproportionately, because assistants preferentially quote sources that answer the question asked.

Give the recommendation somewhere to go. An assistant that names you and then offers a phone number and business hours loses the patient who was researching at 11pm. A real booking path is the difference between a mention and an appointment.

Then maintain it. Citations decay. This is the part that fails in practice, because maintenance has no natural owner in a clinic and no deadline forcing it.

What to be skeptical about

Anyone selling guaranteed AI rankings is selling something that doesn’t exist. There is no ranking to buy and no placement to purchase.

Anyone promising results in a week is wrong. Listing corrections propagate quickly; citation change takes weeks to months.

Anyone recommending mass-generated content is giving you advice that will actively harm you. Volume without substance makes your practice harder to summarize, not easier.

The window

Most independent practices have done none of this. That is unusual and temporary.

The same thing happened with local SEO around 2010 and with Google Business Profiles around 2015: a period where relatively basic, disciplined work produced outsized results, followed by a period where it became table stakes and the advantage evaporated.

Healthcare is early in that curve right now. Which mostly means the practices that start this year will look prescient, and the ones that start in three years will be doing catch-up work against competitors who compounded a head start.

Trendzz handles the whole footprint for clinics — structured facts, listing consistency, question-shaped content, and citation tracking across every major assistant. See how it works.