Clinical reasoning, the capacity to generate and weigh differential diagnoses under uncertainty from incomplete data, is built through sustained cognitive effort: the struggle to interpret ambiguous findings before a correct answer is known. This presentation examines a concern with direct relevance to medical education: generative AI tools that supply likely diagnoses within seconds may remove the very cognitive space in which clinical reasoning is trained, a mechanism this presentation terms the collapse of the productive struggle that expertise depends on.

Drawing on cognitive and educational neuroscience, including research on desirable difficulties, productive failure, and cognitive offloading, the presentation argues that medical students who consistently practice diagnostic reasoning with AI assistance available risk developing what may be described as simulated competence: correct answers produced by the tool rather than by the student's own developing judgment. Early evidence from adjacent educational contexts, showing improved performance with AI assistance but impaired performance once assistance is withdrawn, raises a question medical schools cannot avoid: what standards of evidence should govern the integration of AI into clinical training, given that the outcome in question is the reasoning of the next generation of physicians.

The presentation then extends this argument beyond the medical school. If academic medicine observes this mechanism in its own students, it is well positioned, through its tradition of precautionary standards and evidence-based practice, to contribute a missing perspective to the wider debate on generative AI in children's education, where hundreds of millions of children now use AI tools daily with little long-term neuroscientific evaluation.

The presentation concludes with concrete recommendations for medical curricula: minimum unassisted-reasoning requirements before AI-assisted diagnostic exercises, assessment designs that distinguish genuine clinical judgment from AI-assisted performance, and a call for medical educators to engage actively in AI-education policy discussions beyond their own institutions.