"40 million Americans ask ChatGPT a health question every day": two Harvard doctors describe a parallel care system

In STAT, Arya Rao and Marc Succi argue that announcements about AI that goes beyond doctors have mainly fed private medicine outside the walls, backed by commercial interests.

In a forum published by STAT on August 19, 2026, Arya Rao, a PhD student in medicine and science at Harvard Medical School and MIT, and Marc Succi, a radiologist at Mass General Brigham and associate professor of radiology at Harvard, deal less with model performance than with their effects on the organization of care.

Their starting point is a figure of use: more than 40 million Americans ask ChatGPT a health question every day, most often outside the firm's opening hours. Studies that show models that exceed clinicians on diagnostic reasoning tasks have, in their view, fed an imaginary of substitution that leads to something else: a parallel health circuit, structured by private actors, where the answer refers the user to himself rather than to a caregiver.

The two authors recall that health accounts for nearly a fifth of the U.S. economy, which explains the entry of all major AI companies into this field. Their objection relates to responsibility: this parallel system captures actual clinical issues without providing the monitoring, audit and accountability framework that exists in the hospital.