Your Patients Are Asking AI Before They Ask You

By Mike Woo-Ming, MD, MPH

A patient sits down and says, "So I asked ChatGPT about this and it said it might be my thyroid."

What's your face doing?

Because that's the whole ballgame right there. Whatever you do in the next four seconds sets the tone for the rest of the visit, and probably for whether that patient tells you the truth next time.

You went through this once already. Twenty years ago it was "I looked it up online," and a lot of good doctors answered with an eye-roll and a joke about Dr. Google. What that accomplished was teaching patients not to mention it. They kept searching. They just stopped telling anyone in a white coat.

This time it's bigger, and pretending otherwise doesn't serve you.

Why this one is different from WebMD

Old-school symptom searching gave people a wall of links and a worst-case scenario. It was obviously a search engine and it felt like one.

What patients are using now answers in complete sentences, in a warm tone, with confidence, and it remembers what they said two questions ago. It asks follow-ups. It never seems rushed. It's available at two in the morning, which is when people are most frightened and least able to reach anyone.

That last part is worth sitting with. Some of what patients are getting from these tools isn't medical information at all. It's the experience of being listened to without a clock running. If we're honest about what the fifteen-minute visit has become, that's a gap we left open.

And the output is often reasonable. Not always — it can be confidently wrong, and it doesn't know that the patient left out the medication they're embarrassed about — but reasonable enough that dismissing it outright makes you look like the one who's behind.

The response that actually works

Curiosity. That's it. That's the move.

"Interesting — what did it tell you?"

So do that, and a few things happen at once. You find out what the patient is actually worried about, which is frequently not the thing on the intake form. You get a look at the reasoning they're carrying into the room, including the wrong parts, so you can address them directly instead of guessing at them. And you signal that they can tell you things without being made to feel foolish, which is worth more over the life of that relationship than any single correction.

Then do the thing only you can do. The chatbot had a paragraph of text. You have their chart, their history, their labs from March, their mother's cardiac history, and the fact that they look different than they did in January. Say that out loud. "Here's what that tool couldn't know about you."

That's not a defensive move. It's a demonstration of what you're for.

Where it goes wrong, and what to say about it

Be specific with patients about the failure modes, because vague warnings don't land.

These tools don't examine anyone. They can't hear the murmur, feel the abdomen, or notice that someone's affect is off. They work entirely from what the patient chose to type, which means anything left out — the drinking, the supplement, the symptom that felt too minor to mention — simply doesn't exist in the answer.

They can also be fluently wrong. The tone doesn't change when the accuracy does, which is a genuinely dangerous property. A person who has never seen one hallucinate has no reason to distrust a confident paragraph.

And there's a privacy dimension most patients haven't thought about for a second. Whatever they typed into a consumer app about their health may well be retained on someone's server, under terms almost nobody reads. Your obligations around their information are one thing. What they hand over voluntarily at midnight is another, and they should at least know they did it.

Say all of that plainly, once, in about forty seconds. Most patients take it well, because it's framed as helping them use the thing better rather than telling them to stop.

Get ahead of it in your own practice

If patients are going to consult something before they see you, you have some say in what that something is.

Write the answers to your ten most common questions in your own words and put them where your patients live — your site, your after-visit summary, your email list. Every question you answer well in public is one someone doesn't have to ask a stranger. That's the same reason I keep telling physicians to build owned content, and this is the version of the argument that finally makes it urgent.

Look at your access, too. A meaningful share of AI symptom-checking is happening because there was no faster way to get a human answer. Any improvement to how quickly someone can reach your office reduces the volume of two-a.m. self-diagnosis you'll be untangling later.

And staff will get asked about this. "Should I be using that AI thing?" A one-line practice answer — something like that's not a diagnosis, but bring whatever it told you to your appointment — beats five people improvising five different answers.

Where I actually land on it

My own view is that this is the biggest shift in how medicine gets practiced since the EHR — and unlike the EHR, we still have some say in how it lands.

Patients showing up informed and a little wrong is not a new problem. It's the oldest one we have. The difference is that the informing is better written now.

Your value was never being the only source of medical information. It stopped being that a long time ago. What you have is judgment, examination, continuity, and the ability to look at a specific human being and say here's what this means for you. Nothing I've seen is close to that, and the doctors who handle this well will be the ones patients keep coming back to precisely because the tool sent them.

So when someone opens with "I asked AI about this" — get curious. It's better information about your patient than you'd have gotten otherwise.

Dr. Mike Woo-Ming is a Mayo Clinic-trained physician, practice owner, and founder of BootstrapMD. He's the author of The Positioned Physician: Earn More, Work Smart, Love Medicine Again. Learn more at michaelwooming.com.

This is general educational information, not legal, medical, or compliance advice.