An excerpt from Medicine at the Crossroads: Commentaries from a Profession Caught Between Patients, Politics, Technology, and Trust.
An emergency medicine physician writing in the New York Times recently described an interaction she had with an AI chatbot that might be a teaching lesson for our profession. After receiving her own lab results, she tried to discuss them with her physician. She was told a phone call would require another appointment. So, she did what many patients now do: She turned to ChatGPT.
This wasn’t a case of a patient mistaking a chatbot for a clinician. This was a doctor who understands AI’s limitations, its tendency to hallucinate, to miss context, to sound more certain than it should. She used it anyway. What struck her wasn’t that the chatbot knew more than a physician. It was that it behaved in ways she wished medicine still did. The bot was patient. It asked follow-up questions. It remembered her history. It didn’t feel rushed or dismissive.
ChatGPT didn’t outperform a doctor clinically, as some studies of diagnostic accuracy have shown. It outperformed the system experientially. It won because it had time.
It would be easy to present this as a story about AI encroaching on clinical care, but that really misses the point. What this doctor’s experience highlights is what happens when care is stripped of time, continuity, and relational depth. Patients aren’t turning to AI simply because it’s novel or impressive. They’re turning to it because the alternative, the real doctor, may be inaccessible or hurried, or the visit may feel transactional rather than warm and caring.
The standard response from our profession has been to warn patients about AI. Those warnings are warranted. Generative AI can be wrong. It can miss red flags, overlook details, and present flawed conclusions with unsettling confidence. But if that’s all we say, “Don’t trust the chatbot,” we risk ignoring what patients are telling us by using it.
They want timely answers. They want explanations they can understand. They want the freedom to ask the same question more than once without feeling judged. They want space to process unclear results. In many cases, they want someone, or something, that will simply stay with them long enough to answer questions and settle confusion. The emergency physician, herself, understands this well. She writes, “My experience with the chatbot has already shifted how I interact with patients. I try to listen for what’s behind their questions.”
Survey data on AI health use reinforce this shift. A substantial proportion of Americans now use AI tools for health information, often before or after seeing a physician, or both. This suggests supplementation, not outright replacement. But the reasons are important: Cost barriers, access challenges, and prior experiences of feeling ignored all drive patients toward these tools. That last factor, feeling ignored, should concern us most.
If patients are seeking out AI because they felt unheard, the issue isn’t just technological literacy. It’s relational scarcity. It reflects a system shaped by productivity pressures, administrative burden, and shrinking visit times. AI didn’t create these conditions. It simply stepped into the gap.
The ER doctor had the advantage of clinical training. She could interpret the chatbot’s responses, filter its suggestions, and recognize when to seek care. Most patients don’t have that safety net. That asymmetry is the central challenge of AI in medicine: The tools are widely available, but the ability to use them safely is not. Anxiety following AI health searches is highly common, often leading to a cyberchondria spiral. Many users have found that AI tools are a good place to start and sometimes a dangerous place to finish.
Rather than reacting defensively, clinicians might take a different approach when patients bring AI into the exam room. A patient who says, “I asked ChatGPT about this,” is often trying to engage, not to challenge authority. There’s an opportunity in that moment: What did it tell you? What concerns did it raise? What made sense, and what didn’t? Those conversations can clarify misconceptions and, just as importantly, reveal what the patient is really worried about.
Sometimes the issue isn’t information at all. Repeated questions may reflect worry, confusion, or the need to hear something explained in a different way. AI handles this well because it doesn’t become irritated or impatient, and it often gives slightly different answers to identical questions, allowing patients to revisit the same fear in new language until it feels less overwhelming. But patience and reassurance were never meant to be a machine’s comparative advantage. It’s supposed to be part of the way we deliver care as clinicians.
The risk, however, runs in both directions. Patients may over trust AI because it is conversational and always available. Clinicians, under strain, may begin to rely on it in ways that gradually displace their thinking. As these tools become embedded in documentation, messaging, and clinical decision support, the distinction between assistance and authority becomes critical. If that line blurs, tools designed to support judgment can begin to supplant it.
A more constructive model isn’t physicians versus AI, or patients choosing between them. It’s a three-way partnership. Patients bring their lived experience. Physicians bring judgment, accountability, and medical expertise. AI brings clinical support.
The ER doctor’s experience isn’t a warning about replacement. It’s a reminder. A physician-patient turned to a chatbot because the system made what should have been a simple conversation difficult. The chatbot responded with time, attention, and patience. Those features should not feel revolutionary. They should feel like the practice of medicine.
Arthur Lazarus is a former Doximity Fellow, a member of the editorial board of the American Association for Physician Leadership, and an adjunct professor of psychiatry at the Lewis Katz School of Medicine at Temple University in Philadelphia. He is the author of several books on narrative medicine and the fictional series Real Medicine, Unreal Stories. His latest book is Nobody Told Me There’d Be Days Like These: Hard Truths from Physicians—and What They Mean for Medical Practice.
