Enhancing Problem-Based Learning in Undergraduate Sports Medicine Education with DeepSeek AI: A Comparative Study (Preprint)

BACKGROUND Problem-based learning (PBL) is a cornerstone of modern medical education, yet its effectiveness can be hindered by inconsistent student engagement and facilitator variability. The recent emergence of large language models (LLMs) like DeepSeek offers a potential to revolutionize pedagogical approaches by assisting in content creation and lesson design. OBJECTIVE This study aimed to evaluate the impact of a DeepSeek-assisted PBL teaching model compared to traditional lecture-based teaching on undergraduate medical students’ academic performance, learning efficiency, and satisfaction in a sports medicine course. METHODS A quasi-experimental study was conducted with two randomly assigned classes of fourth-year medical students. The control group (n=33) received a traditional multimedia lecture on sports injuries. The experimental group (n=80) received a PBL session on the same topic, where the content was developed with the assistance of DeepSeek. Both groups completed a 20-item knowledge assessment and a 5-domain satisfaction survey immediately after the 45-minute session. The Mann-Whitney U test was used to compare outcomes due to non-normal data distribution. RESULTS There was no significant difference in post-test scores between the traditional and DeepSeek-assisted PBL groups (82.27 vs 81.00, P=.64). Although the DeepSeek group completed the assessment faster (493.58 seconds vs 556.06 seconds), this difference was not statistically significant (P=.27). Overall satisfaction was high in both groups. Notably, the DeepSeek-assisted PBL group reported higher satisfaction rates in specific domains: 97.5% of students were satisfied with teaching resources (including AI-assisted tools) and instructor interaction. CONCLUSIONS The integration of DeepSeek AI into a PBL framework did not lead to superior academic performance in the short term compared to traditional methods. However, its potential to enhance learning efficiency and student satisfaction with interactive resources is promising. These findings suggest that AI-assisted PBL is a feasible and engaging pedagogical tool. Future research should explore long-term integration, optimized prompt engineering, and qualitative insights to fully leverage AI's potential in medical education.