AI-Facilitated Self-Directed Learning and Mathematics Performance: A Mixed-Methods Study on ChatGPT Use among Generation Z Students

The rapid adoption of generative artificial intelligence (AI) tools such as ChatGPT has raised important questions about how AI-supported learning relates to students’ self-directed learning (SDL) and mathematics achievement, particularly among Generation Z learners. This study examined the relationship between AI-facilitated SDL and mathematics performance using a convergent mixed-methods design among junior high school students in accredited public schools in the Philippines. A total of 272 Grade 7–10 students completed the Self-Rating Scale of Self-Directed Learning (SRSSDL) and an AI/ChatGPT Use and Engagement questionnaire, and their responses were linked to standardized mathematics assessment scores. In addition, 73 students participated in focus group discussions to describe how they used ChatGPT for mathematics learning. Quantitative results showed that most students demonstrated moderate SDL readiness (47.79%), while 29.41% exhibited high SDL and 22.79% showed low SDL. SRSSDL scores were strongly and positively correlated with mathematics performance (r = .849, p < .001), whereas ChatGPT usage showed a weak but significant positive correlation with performance (r = .227, p < .001). SRSSDL and ChatGPT usage were not significantly related (r = .080, p = .187). Multiple regression indicated that SDL readiness and ChatGPT usage significantly predicted mathematics performance (R² = .746), with SDL emerging as the dominant predictor (β = .836, p < .001) and ChatGPT usage contributing a smaller but significant effect (β = .159, p < .001). Qualitative thematic analysis yielded three themes: (1) Personalized support and motivation, (2) Challenges in critical thinking and overreliance, and (3) ChatGPT as a learning companion, indicating that the benefits of ChatGPT depend on how intentionally and critically it is used. Integrated findings suggest that while ChatGPT may provide supplementary support, SDL readiness remains the primary driver of mathematics achievement, and responsible AI use practices are necessary to minimize risks such as overreliance and reduced independent reasoning.