How Good Are Ambient AI Scribe Notes?

Ambient artificial intelligence (AI) scribes eavesdrop on clinical encounters and generate a structured note for the clinician to review and finalize. How good are those notes? In this Veterans Health Administration–led study, researchers audio-recorded 5 simulated primary care visits; then for each, 11 AI scribes from undisclosed vendors and 3 human clinicians generated notes from the recordings. Six blinded raters scored each note using a documentation quality instrument that assesses accuracy, organization, and thoroughness, among other domains (modified PDQI-9). Of note, two visits included accented speech, one had background noise, and in one the patient and clinician wore masks.