When students use AI for clinical reasoning, the obvious move is to ask for the answer. That may be the wrong starting point.
A better role for AI may be to ask questions after the learner has committed to an initial judgement. Used this way, AI does not replace reasoning; it pressures the reasoning to become clearer.
Why answers can arrive too early
Clinical reasoning develops through effort. Students need to notice relevant information, weigh alternatives, manage uncertainty and justify a plan. If AI supplies a polished answer too early, the learner may skip the difficult part.
AI as a questioning partner
After a physiotherapy student proposes an initial hypothesis, AI could ask what finding supports the hypothesis, what finding does not fit, what alternative remains possible, or what red flag would change the plan.
A teaching sequence
A practical sequence is: student reviews the case without AI, records an initial hypothesis, uses AI to generate probing questions only, revises or defends the reasoning, and then receives educator feedback.
The principle
Do not ask AI for the answer first. Ask students for their reasoning first. Then use AI to test that reasoning.