Abstract
<title>Abstract</title> <p>The integration of Generative Artificial Intelligence (GenAI) into programming education has produced a rapidly growing body of empirical research, yet the conditions under which these tools support or undermine student self-regulation remain insufficiently understood. This article presents a systematic literature review on this topic, conducted in accordance with PRISMA 2020 guidelines, encompassing 58 studies published up to October 2025 and retrieved from IEEE Xplore, Scopus, and the ACM Digital Library, complemented by a snowballing procedure. The review investigates how GenAI-based tools and approaches have been integrated into programming education and how this integration influences student self-regulation, operationalized through its manifestations in autonomy, engagement, and motivation. Based on the synthesized evidence, a four-dimensional taxonomy is proposed, organizing the literature along the dimensions of Target Syllabus, Mediation Strategy, Cognitive Impact, and Technical Approach. The proposed taxonomy consists of a descriptive map of the research field that may act as a planning instrument for future pedagogical interventions, contributing to a more systematic understanding of the conditions under which GenAI supports the development of autonomous learners in programming education. The findings indicate that the effects of GenAI on student self-regulation are heterogeneous and are not explained by the technology itself, but by the pedagogical conditions under which it is employed. Structured mediation strategies, including guided use, scaffolded design, and integrated pedagogical ecosystems, are consistently associated with more favorable outcomes, whereas unguided contexts tend to produce superficial interaction patterns and cognitive dependence.</p>