Generative AI can now draft patient explanations, suggest possible differential diagnoses, create clinical cases and provide almost immediate feedback. For physiotherapy educators, the tempting question is whether AI can take over more of the teaching workload.

I think there is a better question: how can we use AI while ensuring that students still learn to observe, reason, communicate and take responsibility?

Physiotherapy is not simply the production of a correct written answer. Students must notice how a person moves, select and adapt an examination, interpret incomplete information, communicate through uncertainty and respond safely to the patient in front of them. An AI system may generate a convincing recommendation without participating in any of these embodied and relational processes.

Promise and caution

My Obsidian notes point to a consistent position: AI use in education should be augmentative rather than substitutive. AI may reduce cognitive and rhetorical workload, widen the questions a student considers and support formative feedback. However, the first AI response should be treated as material for review, not as a finished answer.

Recent work in clinical AI also invites caution. Early studies and preprints suggest that AI-supported systems may help clinicians consider alternatives and generate more personalised rehabilitation material, but they may also omit important context, encourage automation bias and produce patient information that still requires professional correction. Direct evidence within physiotherapy education remains limited.

A reason-first model

This suggests that AI should be introduced as a thinking partner rather than a clinical shortcut. One practical model is simple enough to use in a classroom, simulation session or written case analysis.

  • Reason: the student assesses the case and records an initial judgement before using AI.
  • Consult: the student asks AI for alternatives, questions, counterarguments or feedback.
  • Challenge: the student checks the response against evidence, patient context, professional standards and safety.
  • Decide: the student explains what was accepted, rejected or changed.
  • Reflect: the student identifies what they learned and where uncertainty remains.

What this changes in teaching

This structure makes the student's thinking visible. It changes AI literacy from knowing how to write a prompt into a professional capability involving verification, privacy, bias awareness, disclosure and accountable judgement.

It also separates learning from performance. AI can help a student produce a better-looking answer quickly, but that does not prove that learning has occurred. In clinical education, the developmental struggle matters because students are becoming practitioners who must eventually act without outsourcing responsibility.

  • Use deliberately imperfect AI-generated assessment or treatment plans and ask students to identify unsafe assumptions.
  • Ask students to complete a clinical-reasoning map before consulting AI, then compare what changed and why.
  • Use AI-generated patient education as material for critique, rewriting and person-centred communication.
  • Allow AI to vary simulated patient histories while keeping educator-controlled learning outcomes and safety boundaries.
  • Require students to disclose the tool used, the purpose, the verification process and their own contribution.

Assessment must still protect clinical capability

Assessment should continue to value what AI cannot demonstrate on a student's behalf: observation, hands-on competence, communication, adaptability and defensible clinical reasoning. A student who can paste a plausible AI answer has not necessarily shown that they can examine a patient, respond to distress, recognise a red flag or justify a treatment decision.

The aim is not to protect yesterday's assignments from tomorrow's technology. It is to design learning in which technology strengthens, rather than quietly replaces, the development of a physiotherapist.

Sources informing this article

  • Obsidian note: AI in Physiotherapy Education - Weekly Digest.
  • Obsidian note: Generative AI in Teaching and Learning - PolyU Conference.
  • Obsidian note: Artificial Intelligence - Foundations and Educational Use.
  • Obsidian note: Research Integrity in the Generative AI Era.
  • Obsidian note: Should I Use AI for This?, adapted from Olivier and Rowe's decision guide for doctoral researchers.