This paper presents a systematic review of 55 peer-reviewed studies examining the use of generative AI (GenAI) in promoting physical and mental health among high school and college students. Drawing on the PRISMA framework, we analyze GenAI applications across three domains: personalized wellness planning, early detection of mental and physical health risks, and proactive intervention. The results show that GenAI systems, including large language models and multimodal agents, are increasingly used to generate adaptive fitness routines, support mental health monitoring, and recommend real-time interventions. While the reviewed literature highlights GenAI’s potential to deliver scalable, personalized wellness support, it also reveals critical challenges. These include limited validation in school-based contexts, inconsistent user adherence, over-intervention risks, and unresolved ethical concerns regarding trust, autonomy, and privacy. We synthesize cross-cutting design recommendations for developers and institutions and identify future research opportunities, particularly in longitudinal evaluation, hybrid human-AI systems, and culturally adaptive interfaces. This review contributes a structured foundation for advancing safe, effective, and context-aware GenAI tools in student wellness.