How patients actually find a surgeon in 2026

The path usually starts with a person: a primary care physician, a friend who had the same operation, or an insurance directory that sets the shortlist before anyone else does. Then they look you up before they call. That verification step is now the decisive one, and it runs across search results, reviews, and increasingly an AI answer. The name opens the door. What the patient finds afterward decides whether they walk through it.

That's the whole path in four sentences, and the order matters more than any single channel in it. This piece walks the path as it actually runs, names where practices lose people, and separates what changed from what marketing vocabulary only claims has changed.

Where does the search actually start?

With a person. Referral from a primary care physician, a recommendation from someone who had the same procedure, or a name pulled off an insurance directory because coverage decided the shortlist before anyone else did.

Practices are right about this part: referrals still carry more weight than anything a marketer can buy. A referral arrives with trust already attached, which is why it converts.

What changed is what that name now buys you. A referral used to be close to a decision. Today it's a shortlist entry, because the patient's next move is to check it, and the checking itself surfaces the alternatives.

What happens between the referral and the phone call?

The patient looks you up. That gap between hearing your name and dialing your number is where the modern version of this decision gets made, and nothing in the practice records that it happened.

The check is fast and it's skeptical. Your website, your reviews, your credentials, whatever a search result surfaces, and whether all of it agrees. Patients aren't reading closely at this stage. They're scanning for reasons to stop, and inconsistency hands them one: a suite number that doesn't match, a partner listed who left two years ago, a procedure page that doesn't mention the procedure they were referred for.

A referred patient who can't confirm you online is a patient your referring physician believes they sent you. Nobody reports that loss, because there's nothing to report. The patient booked with the other name.

Do patients really ask AI which surgeon to see?

Enough of them that it belongs in the path. The Pew Research Center's Americans and AI 2026, published 17 June 2026 from a survey of 5,119 US adults fielded 17 to 23 February 2026, found that 60% of US adults say they read AI-generated search summaries, 42% use chatbots to look things up, and 49% have used an AI chatbot at all.

Those figures cover US adults generally rather than surgical patients specifically, so they establish a base rate rather than a fact about your waiting room: reading an AI summary is now ordinary behavior for the majority of adults, and the people choosing a surgeon are drawn from that same population.

Read them carefully, because the interesting part isn't the headline. Patients haven't stopped searching, and they haven't stopped asking their physician. They've added a step, and the added step gets answered with names.

We also checked what the models themselves carry. In July 2026 we put fifteen buyer-intent questions to two AI models with web search switched off, which measures what a model has absorbed rather than what it can look up mid-answer. One of the two models was the wrong instrument, a coding assistant that declined part of the set and hedged much of the rest, so we're not publishing a naming rate for the month. What holds regardless: the brands that came back were patient directories, Healthgrades and WebMD and Vitals among them, and not one of the thirty answers named an individual surgical practice. We publish that measurement in full, with its limits, in what AI models know about finding a surgeon.

Three limits on that, all checkable. Retrieval was off, so it is not what a live assistant tells a patient today. Five live-retrieval surfaces (ChatGPT, Perplexity, Google's AI results, Claude on the web, and Copilot) sit behind a login, so they went unrecorded rather than counted as absent. And the counts move between runs, which is part of why the rate is held: our brand list can only find names it already knows, so treat the pattern rather than any total as the finding. What it does establish is which names are established enough to come back unprompted. Directories are. Practices are not.

There are two distinct moments where an engine can decide something about you. The first is the open question, where a patient with no name yet asks who to see, and the engine's answer is the shortlist. The second is the vetting question, where a patient who already has your name asks what the engine knows about you. The first is the one practices think about when they think about AI at all. The second is the one that touches every referral you already earned.

What patients check when they look you up

Four things, in roughly this order:

• That you're real and current. Address, phone, hospital affiliations, who's still in the group. It reads as trivial until one of them is wrong somewhere public.
• That you do the specific thing they need. Not "orthopedics". The procedure they were referred for, described in words they recognize, ideally on a page of its own.
• What other patients said. Recent reviews weigh more than a high count of old ones, and a thoughtful reply to a critical review does more work than a wall of five stars.
• Whether the story holds together. This is the one practices underestimate. Patients don't audit you. They notice friction, and friction reads as risk when someone is choosing who operates on them.

Notice that none of those four is a marketing asset. They're records. Which is why the practices that do well here are the ones that maintain their public information, and why the fix is a maintenance habit rather than a campaign.

Why do referred patients still disappear?

Because the check failed quietly, and a failed check produces no evidence. There's no missed call, no abandoned form, and no line in any analytics dashboard for a patient who read your page and chose someone else.

Your public record is infrastructure, not marketing. Paid reach can put your name in front of someone who wasn't looking; it does nothing at the moment a motivated patient is checking whether to trust you, and that moment is decided by what is already there.

A profile on a hospital system's site or a large directory can carry a patient through the check too, so it's worth keeping those current. Your own site, your own reviews, and your own consistent record are the parts you can correct the same afternoon.

Working the path in order

Work the path in the order patients move through it, rather than the order vendors sell it.

The referral relationships stay first, because they open the decision and nothing here replaces them. Then make the check survivable: one consistent set of facts everywhere, a real page for each procedure you perform, and current reviews you actually respond to. Then the AI layer, which runs on the same material, since engines quote pages that answer questions and corroborate them against independent sources. Our guide to SEO for doctors covers that shared foundation, and digital marketing for doctors maps how the channels fit together.

The order is the point. A practice that buys ads before fixing its record is paying to send more people into a check it hasn't passed.

You can run the check yourself. Our guide on why AI doesn't recommend your practice walks through the questions to ask and how to read what comes back. Or start from a baseline someone else builds: our free visibility audit shows what patients and engines currently find when they look you up, including which engines mention you today and what they cite. We review every request, and qualifying practices get it back within 24 hours, at no charge.

The check is already happening for any practice a patient has been referred to. The open question is whether you've looked at what it returns.

Correction, 1 August 2026. This page first reported that seventeen of thirty answers named a brand, with per-brand counts. Fifteen of those thirty came from a model that was the wrong tool for the question, so any rate rested on a line drawn by hand between an answer and a decline. The rate is withdrawn rather than restated, and the full account is on what AI models know about finding a surgeon. What did not change: none of the thirty answers named an individual surgical practice.

Common questions

How do patients find a surgeon?

It still starts with a person: a referral from a primary care physician, a recommendation from someone who had the same procedure, or an insurance directory that sets the shortlist. What's changed is what happens next. Patients verify the name online before they call, checking your website, reviews, credentials, and increasingly an AI answer. The referral gets you onto the list; the check decides whether you keep the appointment.

Do patients use ChatGPT to find doctors?

Often enough to plan for. The Pew Research Center's Americans and AI 2026 survey of 5,119 US adults, fielded February 2026, found that 60% say they read AI-generated search summaries and 42% say they use chatbots to look things up. That covers US adults generally rather than surgical patients specifically, so treat it as the base rate the behavior sits on. Patients use it two ways: asking who to see when they have no name, and asking about a specific surgeon whose name they already have.

Are referrals still more important than online search?

Yes. A referral arrives with trust already attached, which is why it converts better than anything bought. The mistake is treating that as a reason to ignore what happens afterward. Referred patients check you before they call, so a weak public record quietly costs you people your referral network already sent. The two aren't competing channels; the second one decides whether the first one pays off.

What do patients look at when they check a surgeon online?

Whether your details are current and consistent, whether you clearly perform the specific procedure they need, what recent reviews say and how you respond to them, and whether everything they find agrees. Patients aren't auditing you. They're scanning for reasons to hesitate, and contradictions between your site, directories, and hospital pages hand them one.

How do you know whether you're losing patients at the check?

You mostly can't see it directly, which is what makes it dangerous. A patient who reads your page and books elsewhere leaves no missed call and no form abandonment. The workable proxies are asking new patients how they found you and what they looked at, and auditing your own public record the way a patient would: search your name, read what comes back, and note every detail that disagrees.

Does a practice need to be on every platform?

No. Consistency across a few surfaces that matter beats presence everywhere. Your own site, your primary review profiles, your professional listings, and the directories your specialty's patients actually use are enough for the check. A dozen half-finished profiles with mismatched details actively hurt, because contradictions are what make a record look unreliable to both patients and engines.