Series: The Patient, the Physician, and the Machine
Part 1 of 3
Perhaps the most humbling realization in medicine is how much of healthcare happens after the doctor’s appointment ends, in the silence our system doesn’t fill.
The visit may be over, but the patient’s uncertainty is not. They still have symptoms, fears, lab results, medication questions, and family members asking what the doctor really meant.
Sometimes a question was genuinely forgotten and comes to mind in the car ride home. Or it shows up at midnight, when a parent is staring at a child’s fever or a patient is rereading a lab result they don’t understand. Many times, patients will nod along with the doctor and affirm they got it, not because they actually do but out of pride. They walk out without understanding and open questions. And increasingly, these questions are being asked to AI.
Consider Nancy. She’s fictional, but every clinician has met a version of her.
Nancy is in her mid-50s, and she’s a middle school teacher in Ohio. After a routine physical, I mentioned that her cholesterol was “elevated” and that she should “watch it.” The appointment ended. She drove home. By the time she pulled into her driveway, she had three questions she’d forgotten to ask. At 6 p.m. that evening, she had left a voicemail on our clinic’s after hours line. By 9 p.m., she was typing them into ChatGPT. Not because she trusts it more than me (I hope). But because I wasn’t there at that moment, and the uncertainty is.
Our current system has created one of the stranger tensions in healthcare right now: people like Nancy are deeply uneasy about AI, but they are still turning to it for medical advice.
A recent KFF poll found that roughly one-third of all adults have used AI chatbots for physical or mental health advice. The same poll found that 77% of the public is concerned about the privacy of personal medical information shared with AI tools.
In other words, millions of patients are worried about AI and using it anyway.
That is a paradox that all of us in healthcare need to sit with for a minute, because it tells us something important. Patients are engaging with a healthcare tool they don’t even fully trust to get answers and reassurance. They want to know whether the thing they are feeling is normal, urgent, dangerous, or worth bringing to a doctor.
This is because the traditional patient journey is still full of gaps: before the visit, after the visit, and everywhere uncertainty lives in between.
OpenAI has reported that more than 40 million people ask ChatGPT health-related questions each day, and that most health-related conversations in the U.S. happen outside typical clinic hours. That makes intuitive sense to anyone who has taken care of patients. Anxiety doesn’t wait for office hours.
So patients do what patients have always done: they look for help wherever they can find it.
Years ago, that meant WebMD or diving into a Reddit rabbithole. Before that, it meant a medical book on the shelf or a neighbor with strong opinions. Today, it means an AI chatbot that answers immediately and in plain language.
Nancy asked ChatGPT what an LDL of 142 means. She asked whether she needed medication, and she asked whether she should change her diet before her follow-up in three months. The AI answered all three questions clearly and without making her feel embarrassed for not already knowing. She went to sleep with less dread than she’d felt in the car. Whether every answer was perfect is a separate issue. Whether she needed someone to talk to that night is not.
It is okay for healthcare professionals to be uncomfortable with that. AI can be wrong, sound confident when it should be cautious, miss nuance or fail to recognize when someone needs urgent care. But if we’re being honest with ourselves, we shouldn’t be surprised.
For years, healthcare has asked patients to navigate complexity with too little guidance. We give them rushed visits, fragmented records, long waits, confusing bills, clinical jargon they may not understand, and portal messages that may or may not get answered in time.
In fact, research shows that many people who use AI for health advice are using it around doctor visits, with 59% researching before seeing a doctor and 56% researching afterward.
We should interpret this as a signal that patients want to understand their care more deeply than our current system allows.
Three months later, right before her next follow-up with me, Nancy used AI again. This time, to prepare. She asked what questions to raise about medications, and about side effects. She walked into that appointment with a written list. I was surprised but happy, because she was one of the most prepared patients I’d seen all year. Nancy didn’t tell me where the preparation came from initially. Only at the end of the visit did she mention she had done homework with AI, as she wasn’t sure I would approve.
AI isn’t the answer to every gap in healthcare. It can be wrong, and in medicine, being wrong can matter a lot. But if we want people to trust the healthcare system more than they trust a chatbot, then we need to build a system that earns that trust.
That means building around one North Star: the patient. Not institutional comfort or knee-jerk measures that blur the lines between control and safety. The patient.
If a tool improves access or supports better clinical decisions, we should ask whether it helps patients. Our instinct should not be to shut it down because it makes the system — or us — uncomfortable.
Of course, oversight and safety matter. But those things should always exist in service of better care, rather than reflexive defenses of the way medicine is currently operating.
The rapid development of AI in healthcare is giving us the opportunity to ask a simple question over and over: does this make care safer, clearer, more accessible, and more humane for patients?
If the answer is yes, we should build around it. If the answer is no, we should fix it.
Fear of change (or anything else) can’t become a substitute for patient-centered judgment. Patients are already using these tools. The only question left for healthcare leaders is whether they will shape how AI is used or watch from the sidelines while patients like Nancy figure it out alone.
Nancy’s story doesn’t end with a medical crisis or a cautionary tale about AI misinformation. It ends the way most patient stories end: quietly, in the in-between spaces the healthcare system wasn’t designed to reach. Nancy is managing her cholesterol, and she is more engaged in her care than she was a year ago. And she still uses AI, because it shows up when nothing else does.
Patients like Nancy are not waiting for healthcare to get comfortable with AI. They are already using it to fill the gaps we have left behind. Our job now is to meet them there, with humility and a system built for the patient.
— Rafid Fadul, MD
