Your Patient Trusts ChatGPT More Than You Now: The New Yorker's Dr. Dhruv Khullar on Medical Authority in the Age of AI

Key Podcast Moments

• The difference between cognitive deskilling and cognitive foreclosure, and why the second is the more alarming threat to physician training
• Why a patient who refused to believe a normal lab result told Graham more about AI and medical authority than almost any research paper has
• What happened when an AI went up against the best diagnostician Dhruv has ever trained alongside, on some of the hardest diagnostic cases in medicine
• How Medicare Advantage insurers turned risk adjustment into a revenue strategy, and what physicians and patients are actually paying for it
• Whether medicine can still be a calling after the system has spent years making it feel like a job

Dhruv Khullar is a hospitalist at New York Presbyterian, a health policy researcher at Weill Cornell, and a contributing writer to The New Yorker. He rounds on patients in the morning and reports on what medicine is becoming in the afternoon. That dual position shapes everything about how he thinks: when Dhruv writes about AI, private equity, or the professional identity of physicians, he is writing as someone who was just in the room.

Dhruv joins Graham this week to talk about what AI is actually doing to clinical reasoning and the authority physicians hold with patients, what private equity ownership and Medicare Advantage gaming have done to the structural integrity of American healthcare, and whether medicine can still function as a calling after the system has spent years making it feel like a job.

The conversation on AI opens with a specific and clarifying problem: patients are increasingly deferring to chatbots over the physicians in front of them, and automation bias is making that deference feel rational even when the AI is wrong. Dhruv draws a careful distinction between cognitive deskilling, the erosion of clinical abilities already acquired by offloading hard thinking to machines, and cognitive foreclosure, the failure of trainees who reach for AI too early to develop that reasoning infrastructure at all. Both risks are real, but foreclosure is the harder one to recover from, and the Cabot demonstration Dhruv witnessed at Harvard, where an AI model matched his best residency classmate on one of medicine's hardest diagnostic case formats in a matter of minutes, suggests the ceiling of what these tools can do is moving faster than training has accounted for.

On the structural state of the profession, Dhruv's frame is precise: dazzling innovation on the surface, structural rot underneath. Private equity now owns roughly 500 hospitals in the United States, and Medicare Advantage insurers have turned risk-adjusted payment into a revenue strategy by capturing diagnostic codes for conditions patients may not know they have and that no physician is treating, because sicker-looking patients generate more money. The value-based care model that Dhruv and Graham both once believed in has not delivered on its promise, and what remains is a profession where most physicians are now employed, treated like any other worker, and operating with a shift-work mentality that is a rational response to the conditions, not a failure of character.

Thank you to our wonderful sponsors for supporting the podcast:

Evidently - Leading AI-powered clinical data intelligence https://evidently.com/

Top Takeaways

Dhruv is not a pessimist, and this is not a dire episode. But what his reporting surfaces, across AI research, payment policy, and the bedside, is a profession navigating several reckonings at once. What follows is his clearest attempt to name them.