Background: Systematically evaluate the effects of structured exercise and behavioral intervention (physical activity [PA] alone/PA + diet) on long-term PA in type 2 diabetes. Methods: Systematic search of 11 databases (inception to March, 2017). Randomized controlled trials investigating structured exercise/behavioral interventions in type 2 diabetes reporting PA outcomes ≥6 months were selected. Results: Among 107,797 citations retrieved, 23 randomized controlled trials (including 18 behavioral programs and 5 structured exercise) met inclusion criteria (n = 9640, 43.6% men, age = 60.0 (4.0) y). All structured exercise trials demonstrated increased objective PA outcomes relative to control (pooling was inappropriate; I 2 = 92%). Of 18 behavioral interventions, 10 increased PA significantly, with effect sizes ranging from 0.2 to 6.6 (pooling was inappropriate; I 2 = 96%). After removing 1 outlier, the remaining 17 studies significantly improved PA (pooled effect size = 0.34), although smaller compared with structured exercise. After removing the outlier, meta-regression also revealed significant direct relationships between total contacts (r = .50, P < .01) and more face-to-face counseling (r = .75, P < .001) and increased PA. However, long-term changes in PA and HbA1c were not related. Conclusion: Both structured exercise and behavioral interventions increased PA in type 2 diabetes, although effect sizes were larger for supervised exercise. The effectiveness of behavioral programs was improved when delivery included more extensive and face-to-face contact.