What Does AI Say About Your Medical Practice?

Try this before you read another word. Open ChatGPT on your phone, sign out of anything that would color the result, and type the kind of thing a nervous new patient types: your specialty and city, or a plain question like which dermatologist near me is good with acne scars, or the name of your practice followed by do they take Blue Cross. Read the answer slowly. Is your practice even in it? Are your hours right? Your services? Does it hand the patient your booking link, or does it walk them straight to the hospital system across town?

Most owners have never done this. They have Googled their own name a hundred times, but they have never asked the machine that a growing share of their future patients are now asking first. And the answer that machine gives is fast becoming the first impression, before your website, before your reviews, before a single phone call.

Patients stopped searching and started asking

This is not a someday trend. It is happening right now, and there are numbers. In the 2026 Patient Choice Report from rater8, a survey of nearly 1,000 US adults released in June 2026, 36 percent of patients who searched for a doctor said AI tools like ChatGPT and Claude influenced their choice. That number sat at 17 percent a year earlier. It has more than doubled, and it now sits ahead of both Google search at 34 percent and personal referrals at 32 percent.

Read that again, because it upends an old belief. For years the safest patient in your pipeline was the referral, the one a friend or another doctor sent your way. Referrals still matter, but as a direct influence on who a patient actually picks, AI has passed them. The person deciding whether to call you increasingly asks a chatbot before they ask a human.

And it is not just the young. The report found the strongest AI adoption in the 45 to 60 age group, at 64 percent. These are people with real conditions, real insurance, and real money, and they are letting an assistant shortlist their doctors.

The scary part is not that AI answers. It is that nobody checks

Here is the finding that should stop every practice owner cold. In that same report, 66 percent of patients said they had, at some point, seen incorrect provider information from an AI tool. Wrong hours, a service the office does not offer, an address that moved years ago. You would think that would make people cautious. It did not. A full 60 percent said they still trusted the AI summary without verifying it further.

Sit with what that means for your schedule. If the AI answer about you is wrong, most patients will not call to double check. They will act on the wrong version. If it says you are closed on Fridays when you are open, they book elsewhere for Friday. If it lists a service you dropped, they show up expecting it. And if it never mentions you at all, you do not lose an argument, you simply never enter the conversation. You lose patients you never knew were looking, and no missed call report will ever show it.

We have written before about how patients form snap judgments from what they see when they Google your practice name. The AI answer is that first impression on fast forward, compressed into one confident paragraph that the patient rarely questions.

Why AI keeps pointing patients to the hospital

Ask an assistant for a specialist in your town and watch how often it reaches for the big hospital system or a national chain. That is not a conspiracy against small practices. It is a data problem, and it is one you can fix.

An AI assistant builds its answer from what it can find and trust across the open web: your Google Business Profile, your website, the directories you appear in, your reviews, and any mention of you in local coverage. Big systems have deep, consistent, well organized information everywhere they exist, so the model has an easy time describing them with confidence. An independent practice with a thin three page website, a half filled profile, and details that disagree from one listing to the next gives the model almost nothing solid to grab. So it plays it safe and names the option it can document. We went deeper on this in why ChatGPT recommends hospitals instead of your practice, and the takeaway is blunt: the model recommends whoever is easiest to describe accurately.

You cannot edit the AI. You can only feed it. It recommends the practice it can describe with confidence, and right now that is usually not the one with the thin website and the listing nobody has touched in two years.

The 15 minute self audit every owner should run

You do not need a tool or an agency to see your exposure. You need a quarter of an hour and the willingness to look. Run these, and read each answer as a first time patient, not as the person who knows the truth already.

• Ask by name. Type your practice name into ChatGPT, Google AI Overviews, and one other assistant, then ask what it knows: hours, address, phone, services, insurance accepted. Note every detail that is wrong or missing.
• Ask by need. Search the way a stranger would, with your specialty plus your city, or a specific problem plus near me. See whether you appear at all, and who appears above you.
• Ask the money question. Try your specialty plus do they take my insurance, or plus how much does it cost. These are the questions that decide a booking, and they are where AI most often guesses wrong.
• Ask for a booking. Try a prompt like book an appointment with a your specialty near your city. Does the answer route toward you, or toward a competitor and a hospital line?
• Check from a clean browser. Do it signed out and in a private window, so you see the neutral answer a real stranger gets, not one softened by your own history.

By the end you will have a short, uncomfortable list. Almost every practice does. The good news is that the same list is your fix list, because everything AI got wrong traces back to a source you actually control.

How to shape the answer AI gives

You cannot rewrite the model. You can rewrite what it reads. AI assistants pull from the public record of your practice, so the whole game is making that record complete, consistent, and clear. A few things move the needle most.

Fix and fill your Google Business Profile. It is the single most quoted source for local provider facts. Correct hours, the right categories, every service listed, current photos, and a steady flow of reviews. A stale profile is the fastest way to get described wrong. Our guide to the Google Business Profile for medical practices walks through the fields that matter, and you can manage it directly through Google Business Profile.

Make every listing agree. When your name, address, phone, and hours match across your site, the directories, and the maps, the model reads that agreement as truth. When they conflict, it either picks wrong or drops you for a source it trusts more. This is old fashioned consistency work, and it matters more than ever, which is why we keep hammering on making sure your practice info matches everywhere online.

Write a website that answers real questions. Assistants love a page that states plainly what you treat, where, for whom, with what insurance, and answers the questions patients actually ask. Clear headings, direct sentences, a real FAQ, and clean structured data let a model lift your facts without guessing. This is the heart of SEO and generative engine optimization for healthcare, and if your site is a thin brochure, that is the first leak to plug.

Keep reviews fresh and human. Recent, specific reviews feed both the ranking and the reputation an AI summarizes. In the same rater8 report, 55 percent of patients said they walked away from at least one doctor because of online reviews, and 75 percent would not book with a provider rated below 4.0. The model reads that signal too.

If you want the technical layer, there is even an emerging convention for handing assistants a clean file of facts about your business, which we covered in whether your practice needs an llms.txt file. It is early, but it points at where this is going: writing not just for the patient, but for the machine reading on their behalf.

This is not a new game, it is the same game with higher stakes

If all of this sounds like SEO wearing a new coat, you are half right. The habits that earn you a top spot on Google, a complete profile, a fast and clear website, consistent listings, and strong reviews, are the exact signals an AI assistant leans on. The difference is the margin for error. A weak online presence used to cost you a few positions on a results page a patient still scrolled. Now it can cost you the entire answer, because the assistant returns one recommendation, not ten, and the patient rarely looks past it.

That is the real shift. Search used to give patients a shelf to browse. AI hands them a single suggestion and most of them take it. Being the second best documented practice in town used to mean second place. Increasingly it means invisible.

Where EtherealMinds fits

We work only with healthcare practices in the United States, and this is one of the fastest changing parts of the job. When a practice comes to us worried that patients are drifting to the hospital or a shinier competitor, one of the first things we do is exactly the audit above: we ask the assistants what they say about you, and we usually find a pile of wrong or missing facts nobody knew were out there.

From there the work is concrete. We tune your healthcare SEO and Google presence so the assistants have clean, consistent facts to quote. We build a website that convinces patients and reads clearly to machines, with the services, locations, insurance, and answers that both need. And we wrap it in a full patient acquisition system, with an AI receptionist that catches the calls those found patients make, so the ones AI finally sends you do not slip away after hours. The point is not to trick the machine. It is to make the true story of your practice easy for it to tell.

The honest takeaway

AI did not ask your permission before it started recommending doctors, and your patients did not wait for you to be ready. A third of them are already letting an assistant shape who they book, and most of them trust its answer without a second thought. That is either a threat or an opening, and which one it is depends entirely on what the answer says. So go ask it. Spend fifteen minutes finding out what AI tells the world about your practice today. Whatever you find, wrong hours, a missing listing, a competitor in your seat, is fixable, and fixing it is just doing the fundamentals well enough that the machine has no choice but to describe you accurately. The practices that win the next few years are not the ones shouting the loudest. They are the ones the machine can describe with confidence.