Patient Experience with Ambient AI Scribes: A Mixed-Methods Quality Improvement Study (Preprint)

BACKGROUND Documentation burden is a leading driver of clinician burnout. Ambient artificial intelligence (AI) scribes have emerged as a tool to reduce this burden by automatically generating clinical documentation from patient encounters. While research has focused largely on clinician outcomes, much less is known about how ambient AI scribes affect the patient experience. OBJECTIVE The objective of our study was to evaluate patient-reported experience during outpatient visits with and without AI scribes. We hypothesized a priori that AI scribe use would be associated with improved provider ratings and likelihood to recommend, due to greater provider attentiveness. METHODS This observational quality improvement study analyzed routinely collected patient experience survey data from outpatient visits across primary care and 21 specialties within a large integrated health system from September 23, 2025, to January 6, 2026. A convenience sample of 38,177 patients who completed post-visit surveys was included, out of which 6,767 encounters involved AI scribe use. Outcomes included patient-reported likelihood to recommend the provider, Net Promoter Score (NPS), ratings of provider listening, explanations, and respect, and qualitative feedback on the AI scribe. Proportional odds and top-box ordinal regression models assessed associations between AI scribe use and patient experience outcomes, adjusting for patient age, race, and visit type. Thematic analysis of free-text responses was performed, and exploratory logistic regression models examined associations between qualitative themes and likelihood to recommend. RESULTS The overall mean age was 62.5 (±17.9) years; 60% were female; 69% of patients identified as White, 28% as Black or African American, and 3% as other races. Patients exposed to AI scribes were more likely to recommend their provider (OR 1.09, 95% CI 1.01-1.19; P=.03), and to give their provider the highest rating for quality of explanations (OR 1.17, 95% CI 1.06-1.28; P=.001), listening (OR 1.11, 95% CI 1.01-1.22; P=.03), and respect (OR 1.16, 95% CI 1.04-1.29; P=.007). Among patients exposed to AI scribes, 91% were willing to use the technology again. Of 1,889 free-text responses, 74.4% were positive, 13.3% neutral, and 12.3% negative. Relational themes (e.g. improved connection and benefit to the provider) were more strongly associated with higher likelihood of recommending the provider, whereas perceived lack of consent and risk of inaccuracy were associated with lower likelihood to recommend. CONCLUSIONS AI scribe use was associated with improvements across several dimensions of patient experience. The elements of patient experience that were most strongly associated with likelihood to recommend the provider were largely relational factors, underscoring the importance of communication, transparency, and clear consent in implementation.