Perspective Raises Concerns About What Ambient AI Scribes Miss

Ambient AI scribes produce clinical notes from encounter audio and are being adopted to reduce documentation burden. This Perspective examines a less-studied question: whether speech-based documentation captures clinicians’ unspoken observations and judgments. It draws on dual-process models of clinical reasoning, research on clinician intuition, and cognitive psychology studies of verbal overshadowing; it does not report a trial of ambient scribes.

The author argues that clinicians may notice features such as appearance, affect, or a sense that a presentation does not fit before they can explain them. Such tacit judgment can prompt further examination or testing, but may be underrepresented in an audio-derived record. The Perspective also raises the possibility that narrating reasoning during an encounter could affect pattern recognition, and that notes may make diagnostic reasoning appear more linear than it was.

The author emphasizes that these concerns are plausible risks, not demonstrated effects: the article states that direct evidence that ambient scribes impair clinical reasoning does not yet exist. It does not recommend abandoning the tools, which address real documentation burdens and have been associated with improvements in clinician well-being. Instead, it calls for evaluation of reasoning fidelity as well as transcription accuracy, and for medical education to distinguish disciplined tacit judgment from bias.