Breakthroughs in artificial intelligence (AI) technology in the field of health management have enabled personalized support for college students’ physical activities. This paper proposes an AI-based personalized exercise prescription system that integrates multi-source data such as physical fitness and health status. A fitness exercise information feedback loop system and a health status assessment model are designed, and a contentbased recommendation algorithm (CB algorithm) is employed to dynamically provide personalized exercise prescriptions for college students with varying health levels and exercise preferences. The study reveals that college students can be clustered into four categories based on their physical fitness levels. The performance evaluation metrics of the model’s exercise prescription system all exceed 0.9, enabling students to achieve exercise effects ranging from 7.682 to 8.606. The six generated exercise prescriptions keep students’ exercise fatigue levels below 12, and students’ physical fitness and exercise skills are significantly superior to pre-experiment levels at the 0.01 significance level.
