Yes — but only if AI is used to create practice, rather than simply provide answers. That distinction is critical.
The easiest application of generative AI in medical education is answering questions. The more interesting application is creating situations in which the student has to answer them.
Don't remove the uncertainty
Clinical reasoning begins with incomplete information. If a system immediately explains the likely diagnosis, differential and management, it may be an excellent tutor — but it has removed much of the reasoning task. Simulation should preserve uncertainty. The student needs to discover what matters.
It can make patients interactive
A simulated patient can respond dynamically to the questions a student actually asks. That creates something static case descriptions cannot easily reproduce: branching discovery driven by the learner. But conversational realism alone isn't enough — the underlying clinical state has to remain coherent.
It can support feedback
Once the encounter finishes, the learner needs to understand what happened. SYNTAX's review model covers history, communication, examination, investigations, differential diagnosis, reasoning, management, safety and professionalism.
That turns an encounter into a learning loop: attempt, evidence, review, reflection, another attempt.
The goal isn't AI practising medicine
The goal is the student practising medicine. That is an important design principle: the system should create the environment, and the learner should do the thinking.