The AI arc and interpretive drift

Artificial intelligence now enters the clinical encounter at three points: before the visit, during clinical reasoning, and after it, when ambient tools generate the note. AI has the potential to adversely influence the diagnostic process at each of these steps. This opinion piece names interpretive drift, the subtle shift in meaning that occurs as a patient's account moves through an AI-generated summary and into the medical record. It argues that reviewing AI-generated documentation is a diagnostic safety practice, not clerical proofreading, and proposes clinical authorship and review as the standard clinicians should apply: active verification that a summary preserved the patient's central concern, distinguished observation from interpretation, and retained meaningful uncertainty. An interpretive frame inherited from a note shapes every subsequent decision, testing, referrals, and treatment, even when the frame itself began with a small distortion. The safeguard is not avoiding AI documentation tools but insisting that the clinician, not the tool, remains the author of the note's clinical meaning. As ambient AI adoption accelerates, training clinicians to critically read AI-generated documentation for interpretive drift should become an explicit, taught skill in clinical education rather than an assumed default.