<p>Intelligent tutoring systems (ITS) are increasingly used to support self-regulated learning (SRL), yet evidence from different educational settings remains scattered. This study reports a meta-analysis of 52 empirical studies that examined the impact of ITS on SRL outcomes. Using random effects models, the analysis yielded a positive overall effect (Hedges‘s g = 0.755), which appears to primarily reflect the combined impact of ITS platforms and specific SRL support features rather than standard ITS alone. Effects differed across outcome categories: skills/competencies and motivation/affect showed larger observed effect sizes (g = 0.816 and 0.800), while the process/behavior category showed a more conservative observed effect (g = 0.585), suggesting the possibility that self-report measures may overestimate ITS effects on SRL. Subgroup analyses showed that several design features systematically influenced these benefits. Hybrid systems integrating multiple intelligent mechanisms showed the largest observed effect (g = 1.295). Systems that embedded regulatory support within subject learning showed larger observed gains (g = 1.048) than systems that offered SRL training alone (g = 0.403). Mixed feedback that combined immediate and delayed feedback showed a larger observed effect than purely immediate or purely delayed feedback (g = 1.554 vs. 0.650 and 0.621). Comparison type, regulation role, education level, disciplinary area, interaction patterns, peer support, and degree of personalization did not show significant moderating effects. These findings suggest that the most effective ITS designs integrate multiple forms of scaffolding, blend complementary feedback, and embed support for self-regulation in meaningful learning activities rather than treating it as a separate add-on.</p>