It is speculated that lifestyle interventions known to improve diabetic metabolic state may exert their effects via adipokines. The aim of this systematic review and meta-analysis was to evaluate the chronic effects of physical exercise on adiponectin and leptin levels in adult prediabetic and diabetic individuals. PubMed, Embase, Scopus, The Cochrane Library, clinicaltrials.gov, and WHO Clinical Trials Registry were searched for randomized controlled trials. Pooled effects of interventions were assessed as mean difference (MD) with random effects model. Sensitivity analysis was conducted to test data robustness and subgroup analysis for study heterogeneity. Twenty-two trials with 2996 individuals were included in the meta-analysis. Physical exercise increased levels of adiponectin (MD: 0.42 µg/mL; 95% confidence interval (CI), 0.23, 0.60, p < 0.00001, n = 19 trials) and reduced leptin levels (MD: −1.89 ng/mL; 95% CI, −2.64, −1.14, p < 0.00001, n = 14 trials). These results were robust and remained significant after sensitivity analysis. Study heterogeneity was generally high. As for physical exercise modalities, aerobic exercise, but not other modalities, increased adiponectin and reduced leptin levels. In conclusion, physical exercise and, specifically, aerobic exercise, leads to higher adiponectin and lower leptin levels in prediabetic and diabetic adults. However, cautious interpretation of current findings is warranted.