AI scribes in medical school: useful teaching tool or a crutch that hurts learning?

AI scribes are software tools that listen to a doctor-patient conversation, or process a clinician's dictated summary, and automatically generate a structured clinical note ready to drop into the electronic health record. Adoption has accelerated sharply in the past two years as hospitals look for ways to cut the hours doctors spend on documentation, widely cited as a leading driver of physician burnout.

The tools are now moving beyond practising physicians and into medical schools and residency programmes, where students and trainees are increasingly exposed to them from their earliest clinical rotations. That shift has opened a debate among medical educators that goes beyond simple efficiency: does learning to write a clinical note with AI assistance from day one still teach the underlying clinical reasoning that note-writing is meant to reinforce?

Proponents argue that the traditional model, in which trainees spend hours transcribing and formatting notes by hand, spends scarce educational time on administrative mechanics rather than on the diagnostic reasoning that actually matters. If an AI tool can produce a serviceable first draft, the argument goes, trainees can redirect that time toward interpreting test results, discussing differential diagnoses with supervisors, or spending more time at the bedside.

Critics counter that the physical and mental act of writing a note is not just administrative busywork. Structuring an assessment and plan forces a clinician to organise their thinking, weigh competing diagnoses, and commit to a clear line of reasoning that can be checked and corrected by supervisors. Skipping that step in training, some educators worry, could produce doctors who are fluent at editing AI-generated text but less practised at generating clinical reasoning from scratch.

Some medical schools have responded by treating AI scribes as a skill to be taught deliberately rather than banned or adopted uncritically. That includes requiring students to draft a note independently before comparing it with an AI-generated version, so they can see where the tool's output diverges from their own reasoning, and requiring supervisors to review both.

There is also a practical safety dimension. AI-generated notes can contain fabricated details, known as hallucinations, or subtly miss important context a clinician provided verbally. Experienced physicians who already know what a good note should contain are generally well placed to catch these errors before signing off. Educators worry that trainees who have not yet built that pattern-recognition skill may be less able to spot an AI-generated error, precisely because they have not yet built up the underlying expertise the tool is meant to support.

Survey data cited by researchers studying the issue suggests that residency programmes vary widely in their policies, with some banning AI scribes outright for trainees below a certain year of training, others allowing unrestricted use, and many having no explicit policy at all, leaving individual attending physicians to set their own rules.

The debate echoes a broader discussion in education about calculators, spellcheckers and now generative AI writing tools: each new technology raises the question of which underlying skill is being automated away, and whether that skill is still worth teaching for its own sake or primarily useful as a stepping stone to a different one.

Medical education researchers say more rigorous studies are needed to determine whether AI-assisted note-writing during training measurably affects trainees' later diagnostic accuracy or clinical reasoning skills, since most of the evidence so far is anecdotal or drawn from small surveys of trainees and faculty.

For now, most educators interviewed on the subject agree on one point: AI scribes are not going away, and medical schools that ignore the technology risk leaving graduates unprepared for how documentation is actually done in modern hospitals. The unresolved question is how to integrate the tools without letting the underlying clinical reasoning skill quietly atrophy.