Two ways of preparing educational presentation about therapeutic exercises to resident physiatrists

This study planned to evaluate the effectiveness of ChatGPT-4o, a generative artificial intelligence platform, in preparing educational slide content on therapeutic exercises for physiatry residents. Both a physiatry expert and ChatGPT-4o developed presentations based on the same reputable sources. Two other physiatry experts, blinded to the origin of the slides and each others’ scores, independently assessed both presentations using the CLEAR criteria—Completeness, Lack of False Information, Appropriateness, and Relevance. Statistical analysis using Cohen's kappa and Kendall's tau-b confirmed significant interrater reliability. The AI-generated slides achieved an overall "good" rating, while the human-prepared slides were rated as "excellent." Although the human-made slides scored significantly higher in individual topics, there was no significant difference when considering the presentations as a whole. These findings support the hypotheses that ChatGPT-4o can prepare effective educational content but that human experts outperform AI due to specialized knowledge. The study suggests that while ChatGPT-4o can assist in creating educational materials, human oversight remains crucial to ensure accuracy and educational value in postgraduate medical training.