Medical training is an apprenticeship based on failure, uncertainty, and the slow accumulation of experience. As AI tools like OpenEvidence become common in clinics, we face a new risk. Trainees who use these tools before building their own internal foundation of clinical reasoning may never learn the core skills required to be independent doctors. The danger is not just that skilled physicians will lose their edge, but that new recruits will fail to develop the ability to reason at all.
Nearly two-thirds of US doctors now use AI for diagnostic support. While these tools offer quick answers based on vast research, they remove the friction that is essential for learning. In the past, a student who struggled to form a differential diagnosis learned from their mistakes. Now, they can receive a near-perfect answer instantly. This makes the trainee look prepared, but it hides a fundamental deficit in their cognitive training.
Medical schools must change how they integrate these tools. We cannot allow AI to replace the mental labor that defines a physician. Supervising doctors should require trainees to perform an unaided assessment before they touch a computer. By forcing a first pass without technology, we ensure that students actually engage in the work of clinical thinking.
This is not a call to reject innovation. Instead, it is a call for discipline. Like pilots who must maintain manual flying skills to handle emergencies, doctors must practice the act of reasoning without assistance. If we do not make this a standard, we will produce a generation of supervisors who cannot catch a machine's errors because they never learned how to find the answer themselves.