Computational Thinking (CT), as a vital component of information literacy and core STEM competencies, has been widely integrated into educational systems globally at all levels. However, how to effectively cultivate students' computational thinking abilities remains a key issue in current educational technology research. In recent years, AI-supported learning has emerged as a significant means to drive innovation in computational thinking instruction. Although existing research has preliminarily revealed the positive impact of AI technology on student learning outcomes, systematic quantitative integration and exploration of moderating mechanisms regarding its intervention effects on computational thinking as a whole and its various dimensions remain lacking. To address this research gap, this study employs a meta-analysis approach to synthesize relevant empirical research published between 2015 and 2025, systematically examining the effects of AI-supported learning on students' overall computational thinking levels and its sub-dimensions. A total of 26 eligible empirical studies were included, yielding 62 effect sizes. Findings indicate that AI-supported learning exerts a moderate positive effect on students' computational thinking (SMD = 0.590). Further moderator analyses reveal that variables such as intervention duration, educational level, teaching strategies, assessment tools and sample size partially moderate the intervention effects. Based on these findings, this study provides theoretical support and empirical evidence for future computational thinking teaching practices integrating AI technology.