Promoting computational thinking (CT) through programming activities is crucial for cultivating talent in today's digital landscape. However, recent research have revealed that students often face significant CT challenges during actual programming tasks. While it is widely recognized that additional scaffolding should be provided to support students' programming activities, there is a lack of comprehensive quantitative synthesis on its impact and how it could function more effectively. The present study adopted a novel three-level meta-analytic method to determine the effectiveness of scaffolding within programming contexts for improving students' CT, as well as to identify key influencing factors. Analyzing 30 empirical studies with a total of 47 effect sizes, the findings demonstrate that: (1) Scaffolding in programming has a high positive effect on students' CT development (Hedges's g = 0.71, p < 0.0001). (2) Intervention duration and educational level emerged as significant moderators, with scaffolding showing greater effects in short-term interventions and among younger students. Based on these findings, several implications for both educational practice and future research have been proposed.