Something changed in the exam room, and most practices have not noticed yet. By the time a new patient books with you, there is a good chance they have already described their symptoms to ChatGPT, read an explanation of what might be wrong, and asked which kind of specialist they need. The research phase that used to happen across a dozen browser tabs now happens inside one conversation with an AI, before your practice is ever in the picture.
I want to be accurate about this rather than sell you a panic. Roughly 32 percent of US adults now use ChatGPT to research symptoms and treatment options, double what it was a year earlier, and the platform fields around 230 million health questions every week. Most people still book through Google and still trust their own doctor, so AI has not replaced provider discovery. What it has done is become the layer that shapes what a patient believes before they choose. And when someone does ask an assistant who the best dermatologist or orthopedic surgeon in their city is, the answer gets assembled from physician profiles, practice websites, review sites, and directory listings. Whether your name is in that answer is not luck. It is something I can work on.
What patients actually do before they book now
When I map how a patient reaches a practice today, the path runs through AI at almost every step. It usually looks like this.
- Research the symptom. They describe what is wrong to ChatGPT and get a plain explanation, often before they have decided to see anyone at all.
- Ask which specialist they need. The assistant tells them whether this is a job for a dermatologist, a rheumatologist, or their primary doctor.
- Ask who is good nearby. They ask for the best provider for their issue in their city, and the assistant names a few.
- Vet a specific name. Once they have a referral or a shortlist, they ask the assistant what it knows about that doctor or practice.
- Read the summarized reviews. Instead of scrolling through fifty reviews, they read the AI summary of what patients say, in the AI's own words.
Notice that four of those five moments happen before the patient ever lands on your website. If your practice is absent or unclear at those moments, you are not losing a ranking. You are losing the patient earlier than any ranking would have reached them.
How AI decides which doctors to name
An assistant has no private opinion about your practice. When it recommends a provider, it is pattern-matching on public evidence, and three signals decide whether you make the list. I treat them as a checklist on every medical project.
- Extractable answers. Your condition and treatment pages state the answer to a real patient question in the first sentence or two, in clean language a model can quote without guessing.
- A single, verifiable identity. Your name, specialty, credentials, location, and hours read the same on the site, on the Google Business Profile, and across the medical directories. Every mismatch is a reason for a careful system to skip you.
- Honest corroboration. Reviews that mention the conditions and procedures you handle, accurate directory listings, and real professional credentials an assistant can confirm somewhere other than your own marketing.
My law background changes how I read this. Google has always been strictest with what it calls Your Money or Your Life topics, and health sits at the center of that category. AI answers inherit the caution. The threshold for being cited in a medical answer is higher than almost anywhere else, which punishes anonymous, generic content and rewards providers who can actually prove who they are and what they do.
The process I run to get a practice cited
There is no schema tag that flips a switch and no way to buy your way into an AI answer. What works is a sequence I run in the same order every time, because each step makes the next one stronger.
- Start with the pages that already rank. They have earned trust. I rewrite their openings so the patient question is answered in the first two sentences, then keep the clinical detail below.
- Turn real patient questions into headings. I pull the actual questions patients ask, from consults, calls, and reviews, and answer each one plainly. Those question headings are what an assistant matches against.
- Fix the identity layer. One consistent name, address, phone, specialty, and provider list across the site, the profile, and the directories, reinforced with Physician, MedicalOrganization, and Person schema so machines read it without ambiguity.
- Prove the expertise. Real, named provider bios with verifiable credentials, board certifications, and hospital affiliations, because an anonymous page of medical advice is exactly what these systems discount.
- Build honest corroboration. A steady flow of genuine reviews and accurate listings on the platforms AI reads, so something beyond your own website confirms your claims.
Answer the patient's real question in the first line
The single highest-leverage change I make is unglamorous. Most treatment pages open with a paragraph about being committed to compassionate, patient-centered care, and a model has to guess what the page is even claiming. When the first line says plainly what the condition is, who the treatment is for, and what to expect, the model can lift that sentence, and lifting it is how you get cited. I write the answer first and let the reassurance follow.
Make your reviews do double duty
Reviews are no longer only social proof for humans. AI systems read them and pull the language patients use straight into their summaries. If your patients keep mentioning a specific procedure or condition, those terms start showing up when an assistant describes you. So I help practices earn reviews steadily rather than in one burst, and reply to each one in a way that stays HIPAA-safe and never confirms anyone as a patient. Recent, specific, well-answered reviews move both your map pack ranking and your AI visibility at once.
Want to know if AI recommends your practice?
I will ask Google's AI Overview, ChatGPT, and Perplexity the questions your patients ask, capture exactly where you are cited and where a competitor is named instead, and send you an honest read on what I would fix first.
Five questions to test this week
You do not need a tool for a first read. You need ten honest minutes and the questions your patients actually ask.
- Write down the five questions a patient asks right before they choose a practice like yours.
- Ask each one in Google and read the AI Overview. Note which practices get named and cited.
- Ask the same questions in ChatGPT and Perplexity, then ask each engine directly what it knows about your practice.
- Mark every question where a competitor appears and you do not. That list is your priority order.
If competitors keep surfacing and your practice keeps being absent, that gap is quietly routing patients elsewhere, and it widens every month it goes unaddressed.
The bar is high, and that favors real practices
It is easy to read all of this as a threat. I see it as the opposite. The systems now standing between you and your next patient reward exactly what a good practice already has: clear answers, a clean and verifiable identity, real credentials, and honest reviews from real patients. The practices that lose are the ones hiding behind anonymous, template copy. If you have built a genuine reputation, most of my job is making that reputation legible to a machine.
Discovery now begins inside ChatGPT and Google's AI, often before a patient has chosen anyone. Being the practice those answers name is the work that maps to your phone ringing, and it is exactly what I get practices cited for.
None of this requires gaming anything. It requires making your credibility easy to read for the patients researching in AI and for the systems summarizing you to them. Do that, and you get found in the old rankings and the new answers at the same time. If you want me to look at where your practice stands today, that is the kind of medical SEO work I do every week.
