The robot doctor is examining the man at the clinic. The future diagnosis of a heart problem.
getty
A recent JAMA perspective by Neal Khosla and colleagues ignited a firestorm of online conversation about whether AI can replace physicians as early as 2030. It is a provocative proposition at a time when healthcare is grappling with workforce shortages and record levels of physician burnout. The authors point to five cognitive tasks for which they argue AI already matches or exceeds physicians’ abilities. AI has already proven useful for many tasks, including gathering patient data, culling guidelines and reducing documentation time. However, the conversation lacks serious consideration of the patient-doctor relationship and whether that matters to people who need care.
It is a common mistake in the tech world to build and plan for the future without fully accounting for human factors like patient preference, predictors of engagement and uptake and access. If you build it, will they come? Will patients even have a choice? Survey data suggest these practicalities are moving in the opposite direction from the technology. A January 2026 survey commissioned by the Ohio State University Wexner Medical Center found that even though 51% of adults reported making an important health decision with AI alone, public openness to AI in healthcare fell from 52% in 2024 to 42% in 2026. This suggests AI replacing doctors is questionable.
The debate rests upon published studies that are largely simulations of single tasks rather than real-time physician encounters with patients in exam rooms, critical care units and surgical suites.
When will AI be ready to eliminate a doctor from the exam room? Nobody knows. Furthermore, are we asking enough questions and the right questions? If AI supplants physicians in five years, will everyone have access or only the people who already own Oura rings and Apple watches? Is healthcare eternally a human endeavor? Should it be? If so, is this journey worth pursuing in this way?
What Are Physicians Saying About AI As The Doctor?
Dr. Roy Lirov is a rare combination, a physician and computer science expert building his own AI infrastructure. He believes AI’s greatest utility in healthcare is supporting clinicians to address administrative burden. “While the pace of improvement is head-spinning, AI replacing docs seems like a bit of a moonshot to me. AI has a long way to go before it is ready,” he says.
His greater concern is whether the field is tracking the benchmarks and measures that matter in clinical care delivery. He says measuring single tasks like creating a differential diagnosis is not a game changer. “Anyone who’s taken care of patients knows exactly how poorly standardized test scores translate to actually practicing medicine. I think we’re fooling ourselves by reporting on what’s easily measured instead of figuring out how to measure what matters.”
He is not alone. Of the thirteen clinicians I asked to react to the notion that AI will eventually replace them, none believed it was possible in the context of medical practice. Their responses did not come from fear or a threat of extinction. Their responses stemmed from accumulated clinical experience, which rarely involves a single, straightforward decision point like the assumptions used to formulate the basis of the JAMA Viewpoint. The practice of medicine is not solely recall, guidelines retrieval and pattern recognition. The practice of medicine is also about intuition, judgment, experience and trust.
Will Patients Accept AI Doctors As The Future?
Ultimately, patients will decide the fate of AI-delivered care. Is it a future they can imagine? What about in rural areas with severe clinician shortages? People already rely on AI chat and search to answer health questions, but over time, will those experiences build enough trust and familiarity to replace a human?
Tamika Williams does not think so. Williams is a community health worker and founder of the Empower CHW Institute. Her experience building trust and helping people in underserved rural communities navigate healthcare shapes her views on AI in healthcare. Despite widespread use of AI-based resources like Claude and AI-powered Google search, she says the tech community has not built trust or familiarity with the communities that stand to gain the most from solving workforce shortages.
“I am pro AI and have used it to help me communicate, but based on what I see and hear in the community, even in the future I can’t imagine how that could work. AI needs to be used in the right context and should not be solely relied on for human relationships,” she says.
Williams is concerned that a message telling people to rely on technology for healthcare engagement would undermine the credibility of community health workers and health institutions in general. “We would have to unlearn everything we have been preaching to communities for decades about seeking a primary care relationship. Why would they trust us then?”
Her comments reinforce the need to incorporate cultural context and access challenges into technology design and development. When communities are omitted from these stages, how can we guard against widening access gaps? If we don’t consider access concerns at the start, are we letting the technology universe decide which communities access state-of-the-art care and which must rely on a chatbot? How can patients and communities have a say in the future of healthcare delivery?
Just Because We Can, Doesn’t Mean We Should
Despite the excitement about the rapid pace of AI in healthcare, in considering the perspectives of its two main stakeholders, clinicians and patients, it seems implausible that physician skills can be completely outsourced to AI. Healthcare is a human endeavor whose foundation is a human-to-human relationship. Who gets to say whether this foundation is no longer valued and necessary? Technologists or patients, families and their clinicians?
In a recent interview with The Economist about AI, Elon Musk said, “There really won’t be anything AI can’t do better than humans, apart from being human.” Humans appear to be capable of building machines that will one day execute most healthcare functions. But if we listen to the wisdom of those delivering and receiving care, a fundamentally human endeavor, just because we can, doesn’t mean we should.