Body movements are regarded as part of the learning process. With the evolution of motion-sensing technology (MST) (e.g., Kinect, Xtion Pro, and Leap Motion), educational researchers try to explore the effect of MST on learning. However, the effect of MST on learning performance is still unclear. This is the first meta-analysis that aims to examine the effectiveness of MST on students' learning. A total of 48 effect sizes from 37 independent and high-quality studies are analyzed, and the result suggests that MST has an upper-medium effect on learning (SMD = 0.574, 95% CI = [0.450, 0.698], p < 0.001), particularly in the affective domain (SMD = 0.822). Moreover, three of eleven moderators (i.e., the subject, site of learning, and region) have moderating effects. The moderator analysis indicates that the following conditions are more conducive to MST-assisted learning: (1) 31[approximately]50 students, (2) middle school, (3) >1 month, (4) No-STEM subjects, (5) game-based learning, (6) small group + individual learning, (7) high embodied level, (8) classroom, and (9) Asia and Europe. Finally, the discussions, implications, limitations, and future research directions are put forward.