peer reviewed · JMIR Perioperative Medicine (2026) · May 27, 2026
AI-Generated Avatar Videos for Postoperative Patient Education Among Health Care Workers: Pilot Randomized Controlled Trial
Syed Ali Haider, Srinivasagam Prabha, Cesar Abraham Gomez-Cabello, Ariana Genovese, Bernardo G Collaco, Nadia Wood, James London, Sanjay Bagaria, Mark A Lifson, Cui Tao, Antonio Jorge Forte
In plain English
Researchers at Mayo Clinic randomised 38 health care workers to receive the same 10 postoperative instructions either as a text handout or as a HeyGen-generated avatar video. Quiz scores were statistically indistinguishable (8.89 vs 8.21, P=.17), but the video group spent nearly twice as long with the material (15.1 vs 8.8 minutes) and rated it significantly clearer, more engaging and better for retention. Participants also complained about unnatural expressions, robotic voices and gestures out of sync with speech — a reminder that engagement gains and comprehension gains are not the same thing, and that this pilot measured the first, not the second.
Related research
- Learning Gain and User Experience of AI Avatar-Based and Human-Presented Explainer Videos: Prospective Randomized Crossover Feasibility Study
- Video Lectures With AI-Generated Instructors: Low Video Engagement, Same Performance as Human Instructors
- Learning Gain and User Experience of AI Avatar–Based and Human-Presented Explainer Videos: Prospective Randomized Crossover Feasibility Study