Why Patients Still Prefer Doctors Over AI Medical Advice

When it comes to health questions, most patients would rather wait than get an instant answer from a machine. A new survey of more than eleven hundred patients found that a clear majority prefer a clinician’s review over unreviewed AI Medical Advice — and women and Gen Z respondents feel most strongly of all.

The study, carried out by researchers from Columbia University School of Nursing in New York and reported by Nursing Times this week, asked a cross-section of patients of different ages and ethnic backgrounds a simple question about hospital portal messages: would you rather have an answer from artificial intelligence within five minutes, or wait up to two days for a doctor or nurse to review and edit that answer? For medical messages — symptoms, treatment questions, clinical concerns — nearly two in three chose to wait for the human.

The trade-off patients were offered

The researchers tested three kinds of messages. For logistical questions such as appointment scheduling and prescription refills, just over half of participants said they preferred the faster AI response. For general, non-medical questions, opinion split evenly. But the moment the topic turned medical, the preference flipped hard: 63% said they would rather wait for a health professional to review the message than receive an unreviewed AI answer.

The generational split is the detail hospitals should notice. Younger patients — the demographic most comfortable with chatbots in every other part of life — showed an even stronger preference for human review than older respondents. The finding cuts against the assumption that digital natives will happily accept automated answers everywhere. When the stakes are their bodies, they want a person.

Natalie Benda, an assistant professor at the nursing school and lead author of the study, said the results confirm a hunch clinicians have held for years. “When people have a medical question, they want to hear from their doctor, not AI, even if AI may give an immediate answer.”

Where AI does get a yes

Patients are not rejecting the technology outright — they are drawing a line around it. AI is welcome for the boring stuff: booking, refills, admin. It is the diagnosis-and-treatment zone where trust collapses. The survey’s mental-health section tells the same story in starker terms. Asked how likely they would be to use support from a licensed human therapist available in three months, an AI chat monitored by a therapist available in one month, or an AI chat alone available immediately, respondents picked the human first by a wide margin. Nearly two-thirds said they would take the human therapist despite the long wait, while only about four in ten would use the AI chat alone with no wait at all.

That hierarchy — human first, human-supervised AI second, unsupervised AI last — is a clear product roadmap for health systems racing to deploy chatbots. Speed is not the selling point patients want; supervision is. A hybrid model, in which AI drafts the message and a clinician signs off, matches almost exactly what respondents described as acceptable. Visible human oversight is what makes AI Medical Advice trustworthy to them.

What hospitals should take from this

The timing matters. Health portals are rapidly adding AI-drafted replies to cope with message volumes, and several systems now let algorithms answer patients directly. This research suggests that rolling out unreviewed AI Medical Advice for clinical questions could erode the trust portals were built to create — especially among the young women who use these systems most. Transparency may be the fix: telling patients when a message was AI-drafted and reviewed, versus written by their doctor, lets them calibrate their trust.

There is also a lesson for the AI industry’s favourite metric. In consumer tech, shaving seconds off a response is a win; in medicine, patients will happily trade two days of waiting for the confidence that a trained professional read their message. As new research on brain health and major reviews of sleep and stroke risk keep reminding us, medical decisions carry consequences that no chatbot bears. And for families tracking the youngest patients, new tools for monitoring children’s screen time show the same instinct: when health is on the line, people want oversight they can trust.

The bottom line is reassuringly old-fashioned. Patients see real value in AI for scheduling and paperwork, but for the questions that keep them up at night, they still want a human being on the other end of the message — even if it means waiting.