Abstract Background Self-efficacy is an important determinant of physical activity but it is unclear how best to increase self-efficacy for physical activity and to maintain these changes. Purpose This systematic review aimed to identify which specific behavior change techniques (BCTs), BCT clusters, and number of BCTs were associated with changes in postintervention and maintained changes in self-efficacy for physical activity across all adult populations. Methods A systematic search yielded 180 randomized trials (204 comparisons) which reported changes in self-efficacy. BCTs were coded using the BCT Taxonomy v1. Hierarchical cluster analysis explored the clustering of BCTs. Meta-analyses and moderator analyses examined whether the presence and absence of individual BCTs in interventions were associated with effect-size changes for self-efficacy. Results Small intervention effects were found for postintervention self-efficacy for physical activity (d = 0.26; 95% CI: [0.21, 0.31]; I2 = 75.8 per cent). “Information about social, environmental, and emotional consequences” was associated with higher effect sizes, whereas “social support (practical)” was associated with lower effect sizes. Small and nonsignificant effects were found for maintained changes in self-efficacy for physical activity (d = 0.08; CI: [−0.05, 0.21]; I2 = 83.8 per cent). Lack of meaningful clustering of BCTs was found. A significant positive relationship was found between number of BCTs and effect sizes for maintained changes in self-efficacy for physical activity. Conclusions There does not appear to be a single effective approach to change self-efficacy for physical activity in all adults: different approaches are required for different populations. Interventions with more BCTs seem more effective at maintaining changes in self-efficacy for physical activity.