Integrating generative artificial intelligence into medical education: effects of prompt instruction on self-directed learning

Generative Artificial Intelligence (GenAI) is increasingly used to support self-directed learning (SDL) in medical education, but limited research has examined how prompt instruction structures medical students’ interactions with GenAI. This study integrated GenAI into Sawatzky’s SDL framework and examined the effects of prompt instruction on SDL processes and learning outcomes. A 10-week quasi-experimental study was conducted with 306 medical students. Learning outcomes were assessed through pre- and post-tests, SDL patterns were examined using ENA of reflective journals collected at two time points, and students’ perceptions were explored through thematic analysis of semi-structured interviews. Students receiving prompt instruction (N = 122) outperformed those without such instruction (N = 184) on the knowledge test. Prompt instruction also shaped students’ SDL process by guiding them to identify knowledge gaps and break down tasks before interacting with GenAI, whereas those without such instruction relied more on repeated revision of generated responses. These differences in interaction structure persisted over time. Interviews further showed that prompt instruction helped students view GenAI as a tool for reasoning and information verification. These findings suggest that prompt instruction can improve learning outcomes and structure SDL in GenAI-assisted medical education.