Impact of Generative AI on Self-Regulated Learning and Cognitive Offloading

Generative artificial intelligence (AI) is rapidly reshaping how senior high school students study, yet its implications for self-regulated learning and learning durability remain poorly resolved in authentic classrooms. This study examined whether a structured generative AI approach strengthens self-regulated learning while increasing cognitive offloading, and whether offloading predicts weaker delayed retention. Using a cluster randomised pretest and posttest design, intact class sections from one public and one private senior high school in Davao City, Philippines were assigned either to structured generative AI use guided by plan, monitor, and evaluate prompts or to non-AI study support. Self-regulated learning was measured through adapted MSLQ strategy scales, cognitive offloading through a brief offloading tendency scale, and achievement through unit quizzes including a two-week delayed retention assessment. Mixed-effects models accounted for section-level clustering and adjusted for baseline scores and access indicators. Structured generative AI use was associated with higher posttest self-regulated learning and substantially higher cognitive offloading. Offloading was negatively associated with delayed retention after adjustment, while the direct condition effect on delayed retention was not reliable. These findings underscore the need to integrate generative AI as a scaffold for verification and metacognitive control rather than as a substitute for reasoning.