A dermatologist we talked to this fall did something a lot of owners are doing lately. She opened ChatGPT and typed a plain question: who is a good dermatologist in her city for acne. She has been in practice eleven years, has great reviews, and a loaded schedule. The answer named the big hospital system across town and two larger groups. Her practice, two blocks from one of those groups, was nowhere in it.
Her first reaction was the normal one. The thing is broken, or it is biased toward the big names. Both of those are partly true. But the more useful reaction is a question: what is the AI actually reading when it builds that answer? Because once you know what it reads, you stop guessing and start fixing the right things.
We dug into the most recent research on exactly this, and the picture is clearer than you might expect. There are a handful of sources these tools lean on, and most independent practices are barely present in any of them.
First, patients really are doing this now
It is tempting to wave this off as something only tech early adopters do. The numbers say otherwise. Reporting in Medical Economics in 2026 noted that among patients who looked for a new doctor in the past year, AI tools were named a top influence by about 36 percent, edging out plain Google results at 34 percent and personal referrals at 32 percent. Separate research found more than one in four patients now use tools like ChatGPT, Google AI Overviews or Perplexity when they research a procedure or build a shortlist of clinics.
So the dermatologist's little experiment was not a party trick. It was a preview of how a real slice of her future patients are already choosing. If you are invisible in that answer, you are not losing an argument, you are losing the appointment before the patient even knows your name.
The surprising part: AI does not read your Google listing
Here is the finding that makes most owners sit up. Promptwatch analyzed the citations behind ChatGPT answers for a batch of organic healthcare questions, covering treatments, medications, insurance and finding a doctor, over a stretch from early September to early October 2026. They looked at which websites the model actually pulled from to build its answers (Promptwatch, 2026).
Five kinds of sources carried most of the weight. Specialty association and credentialing pages alone accounted for roughly 22 percent of the citations. State medical board records made up about 5 percent. Review sites, the ones most owners obsess over, came in under 4 percent. And here is the kicker: not a single citation pointed to a Google Business Profile.
Read that again, because it reshapes how you think about this. Your Google Business Profile is still the engine of Google Maps and the local pack, and it still matters enormously there. But when a patient asks an AI chat tool to recommend a doctor, that beautifully optimized Google listing is mostly sitting on the sidelines. The AI is reading somewhere else.
The five sources AI actually reads
Put the research together and a short, specific list emerges. These are the places AI tends to pull from when it decides whose name to say. Medical Economics framed it well in a piece literally titled audit these five sources before you pay for AI visibility. Here they are in plain terms.
1. Your own practice website
This is the anchor. The AI reads your site to learn who you are, what you treat, where you are located and what makes you credible. If your website is thin, vague or buries the basics inside images and sliders the machine cannot read, you have handed it almost nothing to work with. A clear, fast, well built website that converts is not just for patients anymore. It is the primary document the AI reads about you, and it needs your specialty, your conditions treated, your location and your credentials written in plain, crawlable text.
2. Specialty board and association profiles
Credentialing and association pages were the single largest slice of citations in the research, above 22 percent. These are your profiles on your specialty board, your academy or society, and similar professional bodies. Many owners claimed these years ago and never looked again, so the photo is old, the address is wrong, or half the fields are blank. To an AI judging credibility, a complete, current credentialing profile is gold. A stale one is a missed signal.
3. State license records
Your state medical or dental board license record is one of the most trusted facts about you on the entire internet, which is exactly why AI leans on it. It confirms you are real, licensed and in good standing. You cannot do much creative marketing here, but you can make sure your name, practice and location are accurate and match everything else. Mismatches between your license record and your other listings slowly chip away at the trust the machine places in you.
4. Regional Top Doctors lists and local roundups
Editorial lists, think regional Top Doctors features, best of the city roundups and reputable local directories, show up often in AI answers because they read like independent, human judgment. If you qualify for a legitimate one, being on it is a strong vote in your favor. The key word is legitimate. We are talking about real editorial recognition, not pay to play badges that add nothing.
5. Hospital and academic rosters
If you have any affiliation, admitting privileges, a teaching role, a residency connection, those rosters carry weight. This is part of why big systems dominate AI answers: their rosters are deep, well linked and constantly crawled. An independent practice with even one genuine affiliation should make sure it is listed and spelled correctly, because that single mention ties you to a source the AI already trusts.
Why hospitals keep winning the AI answer
Once you see the list, the dermatologist's experience makes sense. A hospital system has a deep website, a full provider directory, academic rosters, association ties and a trail of mentions across authority sites. It gives AI a mountain of consistent, trustworthy text to stand on. The independent practice down the street often has a single thin website and almost nothing on the credentialing and editorial pages the AI weighs most.
It is not that the hospital is better at treating acne. It is that the hospital is easier for the machine to verify and describe. We went deeper on this specific problem in why ChatGPT favors hospitals over independent practices, and the core fix is the same: give the AI more trustworthy, consistent material to read about you.
How bad is the gap? One study of 216 health practices in New Jersey, run across ChatGPT, Claude and Gemini, found that most independent local practices were essentially invisible in AI generated recommendations. Not ranked low. Not even mentioned. That is the silent cost of ignoring these sources while a competitor tends to them.
What this means for your practice
The good news hiding in all this is that the fixes are concrete, not mysterious. You do not need to crack a secret algorithm. You need to be present, accurate and consistent in the handful of places AI already trusts. Here is the audit we walk clients through.
- Read your website like a stranger. Does it say, in plain text, your specialty, the conditions and treatments you handle, your city and your credentials? If a human has to hunt for it, the AI never finds it.
- Claim and finish every credentialing profile. Specialty board, academy, society. Fill in every field, fix the old details, add a real bio.
- Verify your state license record. Confirm it is current and that your name and location match your other listings exactly.
- Pursue legitimate editorial recognition. If you qualify for a real regional Top Doctors or best of list, get on it.
- Make every mention consistent. Same name, same address, same phone, everywhere the AI might read you.
That last point connects to something we harp on constantly, because it is the cheapest win in the whole list. When your details conflict across the web, both search engines and AI lose confidence in you. We covered the full version in why your practice info has to match everywhere online.
Where this fits in the bigger search picture
It is worth being honest about one thing: this is not a brand new game, it is a new front in an old one. The work that makes you show up in AI answers, a strong website, consistent credible mentions, real authority, is the same work that has always built local search visibility. The question of whether this is truly separate from classic SEO is a fair one, and we laid out our honest take in SEO and AI search for healthcare in 2026.
You also should not swing the other way and abandon Google because one study found no AI citations to Google listings. Patients still open Maps, still read the local pack, still judge you by your reviews. The map and the chat answer are two doors into your practice, and right now most owners have polished one and ignored the other. The winners over the next year will be the ones who tend to both.
How EtherealMinds handles this for practices
We only work with healthcare practices in the United States, so this is squarely our world. When we build a practice's visibility, we treat the AI answer and the Google result as one connected job, not two hobbies. That means a website written so both patients and machines can understand exactly who you are and what you treat, credentialing and association profiles claimed and completed, license and directory details made consistent, and the kind of authority building that earns real mentions over time. It is the steady foundation underneath our healthcare SEO work, and it feeds the same patient acquisition system that turns visibility into booked appointments.
We also will not sell you a magic AI visibility gadget before the basics are right. As the Medical Economics piece put it, audit the five sources first. Most practices find they can move the needle a long way just by fixing what is already broken or blank.
The bottom line
When an AI recommends a doctor, it is not reading your Google listing or your star count. It is reading your website, your credentialing and association profiles, your license record, the editorial lists you appear on, and the hospital or academic rosters you are tied to. Most independent practices are thin or absent across all five, which is why the big systems keep getting named. The fix is not a trick. It is showing up, accurately and consistently, in the exact places the machine already trusts. Do that, and the next patient who asks an AI for a good doctor in your city finally hears your name.
Want to know if AI can even find your practice?
Book a free strategy call. We will check how your practice shows up across the sources AI actually reads, flag what is thin, missing or inconsistent, and show you exactly what to fix first. Built only for United States healthcare practices.
Book a free strategy call →