How does GenAI support self-regulated learning? Evidence from learners' questioning behavior analysis

Generative AI (GenAI), such as ChatGPT, provides an instant and individualized learning environment that has the potential to enhance students’ self-regulated learning (SRL) compared to traditional SRL interventions. However, the impact of GenAI on student SRL remains unclear, particularly regarding how educators can best support self-regulation in GenAI-enhanced learning environments. To address this question, this study conducted an 8-week longitudinal experiment with 30 Chinese university students, combining pre- and post-test SRL questionnaires with a systematic analysis of 786 learner-GenAI dialogs. Through rigorous coding, interactions were classified into six question types (Remember, Prompt, Analysis, Reflective, Apply, and Irrelevant), mapped to cognitive, metacognitive, and motivational SRL strategies, to analyze GenAI’s impact through question type distributions, temporal pattern evolution, and cross-group proficiency comparisons. The results indicated that GenAI was effective in promoting cognitive strategy development, and that learners demonstrated increasingly sophisticated questioning behavior. However, support for metacognition and motivational regulation was limited, and the distribution of question types indicated uneven development between cognitive and metacognitive strategies. Cross-group analysis further suggested that GenAI might inadvertently widen SRL skill disparities without targeted scaffolding. These findings underscore the need for balanced intervention strategies and adaptive support tools to optimize SRL development in GenAI environments.