There is evidence that physical activity positively influences cognition and academic outcomes in childhood. This systematic review used a three-level meta-analytic approach, which handles nested effect sizes, to assess the impact of physical activity interventions. Ninety-two randomised control trials in typically developing children (5–12 years old, N = 25,334) were identified. Control group type and intervention characteristics including duration, frequency, and teacher qualification were explored as potential moderators. Results showed physical activity interventions improved on-task behaviour with a large effect size (g = 1.04, p = 0.03 (95% CI: 0.08–2.00), very low-certainty evidence) and led to moderate improvements in creativity (g = 0.70, p < 0.01 (0.20–1.20), low-certainty evidence). Small beneficial effects were found for fluid intelligence (g = 0.16, p = 0.03 (0.02, 0.30), moderate-certainty evidence) and working memory (g = 0.18, p = 0.01 (0.07–0.29), very low-certainty evidence), but no overall benefit was observed for attention, inhibitory control, planning, cognitive flexibility or academic outcomes. Heterogeneity was high, and moderator analyses indicated beneficial effects of physical activity (PA) with academic instruction of 6–10-week duration with moderate or moderate to vigorous intensity on mathematics outcomes and enriched PA programmes on language outcomes. In contrast, aerobic PA with moderate to vigorous intensity benefitted executive function outcomes. These results therefore suggest differential mechanisms of impact of different types of PA on different aspects of cognition.