BMJ Open Quality (12/22/25) Artsi, Yaara; Sorin, Vera; Glicksberg, Benjamin; et al.

With the emergence of artificial intelligence (AI)-generated text-to-video tools in medical and patient education, a review of five empirical studies shows that evidence for their accuracy, effectiveness and safety remains limited and inconsistent. The included studies — spanning ophthalmology, plastic surgery, dysphagia rehabilitation and neurosurgical training — reported mixed results. Traditional materials were more accurate than AI-generated videos in ophthalmology, while an AI-assisted video game improved swallowing outcomes in dysphagia and patients preferred a video-avatar tool over a text chatbot in a plastic-surgery context. Small sample sizes, heterogeneous outcome measures and limited reporting of confidence intervals prevented meta-analysis. Overall, AI-generated videos may boost engagement or support specific outcomes, but accuracy concerns and methodological variability mean they should currently serve only as adjuncts to established educational resources until robust evidence demonstrates non-inferiority on key outcomes.

Read More