Your Patients Are Asking ChatGPT Before They Call You. What That Quietly Changed About Medical Marketing.

A patient shows up already convinced she has something

She had symptoms for two weeks. Fatigue, a rash that came and went, joint stiffness in the morning. Before she called any practice, she opened ChatGPT and typed a paragraph.

The model gave her three possibilities, a short explanation of each, and a line about when to "seek medical attention." It also mentioned two specialties she might consider. She read it twice.

Only then did she open Google and type the name of a clinic a friend recommended. She skimmed the reviews, did not read the website, and booked an appointment through the phone number at the top of the map pack.

When she walked in, she already had a differential in her head. She was polite about it, but you could tell she was testing what the doctor would say against what the chatbot had said.

This is not a rare story anymore. It is most first appointments in 2026.

What actually changed in the patient journey

For about fifteen years, the patient journey looked roughly the same: symptom, Google search, read a few websites, maybe check reviews, call.

The shape of it now:

Symptom → large language model → maybe Google, maybe not → reviews → call.

The model does the work that websites used to do. It summarizes conditions, explains treatment options, lists specialties, and sometimes even recommends specific providers by name. Healthcare Brew reported in February 2026 that health systems are now competing directly with AI search tools for patient attention, not just with each other.

Google did not stand still. BrightEdge's healthcare tracking shows that AI Overview presence in health-related searches climbed from about 59 percent of queries to nearly 89 percent in two years. Treatment questions, symptom questions, and pain-related questions now almost always return a synthesized answer at the top of the page. The ten blue links are still there. Most patients do not scroll to them.

So the marketing your agency was doing in 2022 and 2023 was optimizing for a step in the journey that, for a growing share of patients, no longer exists.

The three specific things that quietly broke

I want to be specific, because this kind of piece tends to drift into "AI changes everything" mush.

Three things broke in medical marketing, and most practices have not noticed because their monthly call volume is still okay. That is the problem. By the time call volume drops, you are twelve months behind the curve.

1. Your website stopped being the first impression

When a model summarizes a treatment and names your practice as one possible option, the patient forms an impression before your homepage loads. That impression is built from your reviews, your Google Business profile, the way your name is described in articles and directories, and whether your site's content was clear enough for the model to extract in the first place.

If your website exists to make a first impression, you are late. The first impression was already made, somewhere you cannot see.

2. The content you paid for may be invisible

Here is a specific pattern I see constantly when I audit clinics: the site has twenty or thirty blog posts written by an SEO agency three years ago. The posts are 600 to 900 words each. Every one is structured the same way: "What is X? Who needs X? Benefits of X. Contact us."

Those posts ranked fine in 2021. They do almost nothing in 2026. A language model cannot extract useful, specific claims from them because they do not contain any. The AI Overview skips them. ChatGPT skips them. The patient never sees them.

This is not a theoretical loss. It is paid work that is quietly underperforming while your agency reports on "content produced."

3. Reviews became citations, not just social proof

Reviews used to reassure humans. They still do that. But they now also feed the models. When a patient asks ChatGPT or Perplexity for "the best oculoplastic surgeon near me," the model's opinion is built in part from review text, review volume, and how consistent the praise is across platforms.

One Reddit-documented dental case I keep thinking about: a practice with 4.9 stars and 800 reviews was losing to a competitor with 4.7 stars and 2,100 reviews, even though the 4.9-star practice had better outcomes. The model was not reading medical outcomes. It was reading review density.

Why most agencies will not tell you this

There is an uncomfortable reason most medical marketing agencies are slow to acknowledge what is happening: it invalidates a lot of what they sell.

If the patient is informed before they reach your website, conversion rate optimization on your homepage is a smaller lever than it was. If the patient chooses from a set of three names surfaced by an AI, visibility inside that retrieval layer matters more than position three on Google. If reviews function as citations in a model's answer, review strategy is no longer a "nice to have" line item.

Agencies that built their service around ranking reports, monthly blog posts, and clicks are not going to lead this conversation. You will have to.

What actually works in this new shape

A few things, specifically, and honestly some of them are boring.

Structured, specific content. Not "What is Invisalign," but a page that says exactly which cases you treat, what you charge, what recovery looks like, what kinds of patients are not good candidates, and how your approach differs. Models extract specifics. They skip generalities.

Reviews with substance. A review that says "great experience" is worth less to the retrieval layer than a review that says "I went in for a consult about melasma after three months of steroid creams failing, and Dr. Ramirez walked me through a combination of tranexamic acid mesotherapy and pigmented laser." The second one contains named treatments, named conditions, named protocols. It shows up in the model's answer.

An accurate Google Business profile. This is still the single most neglected asset I see. Practices will spend $4,000 a month on ads while their GBP has last year's hours, no services listed, and five unanswered questions from patients.

A response system that actually answers. When a patient does get to the phone call step, they have already waited through three funnels. If you make them wait again, you lose them. We have measured this over and over on our case studies page — the practices that fix response time see booked rates jump without touching ad spend. Simetris went from steady bookings to a 308x ROAS quarter. That did not come from better ads. It came from the gap between the click and the booked appointment finally closing.

A site written for both the patient and the model. That means clear H2s, FAQ blocks, specific numbers, schema markup that actually describes what the page is about. Our medical SEO page walks through this, but even without us, you can start by checking whether your homepage mentions, in plain English, what conditions you treat.

The test that tells you where you really stand

Before you change anything, run this six-point self-audit. It will not take more than an hour.

1. Open ChatGPT (or Claude, or Gemini). Ask it: "What is the best [your specialty] in [your city]?" If your practice does not appear in the response, that is a signal, not a verdict, but it is a signal.

2. Ask it one of your most common patient questions. For a dental practice, "How much does an Invisalign treatment cost in [city]?" For a dermatology practice, "What is the best treatment for melasma in adults?" See whether any specific, named practice comes up. Then check whether your content could plausibly have fed that answer.

3. Google your top three services. Read the AI Overview that appears. Does it cite any local practice? Are you one of them?

4. Open your Google Business profile. Check hours, services, photo recency, unanswered questions, response rate to reviews. If anything is out of date, your retrieval layer is leaking.

5. Read three of your reviews out loud. Do they contain specifics, or do they say "great staff, highly recommend"? If the latter, your review strategy needs work.

6. Call your own practice at 7:15 pm. See what happens. Then call it at 12:45 pm on a Saturday. The answer to both is often the same, and often bad.

If more than three of these came back uncomfortable, you do not have a marketing problem. You have a discoverability-and-response problem. Those are different things, and most agencies are equipped to address neither.

The honest caveat

I want to acknowledge something: not every specialty is feeling this shift at the same pace. High-trust, high-referral specialties — certain surgical oncology practices, some reproductive medicine clinics — still run most of their growth on direct referrals and physician networks. For them, the AI search shift matters less right now.

For everyone else — dental, dermatology, aesthetic medicine, primary care, psychology, med spa, urgent care, general pediatrics — it matters a lot, and it is matter more every quarter.

Where to go from here

If the audit came back uncomfortable, the next step is not to panic-rebuild your site. It is to look at the whole acquisition stack honestly and figure out which piece is actually leaking.

We help practices do this with a free growth diagnosis. It is not a sales pitch. It is a 30-minute review where we look at ads, SEO, AI visibility, response time, and CRM attribution and tell you which one is costing you the most patients right now.

The one thing I would not do is wait another quarter to find out. Patient behavior already changed. The practices that noticed first are the ones compounding.

— Hector Arriola, Founder & CEO, Hillflare