Purpose: The effect of physical activity on glycemic variability remains controversial. This meta-analysis aimed to assess the overall effect of physical activity treatment on glycemic variability in patients with diabetes. Methods: PubMed/MEDLINE, Embase, and Cochrane databases were searched for clinical trials that conducted in patients with type 1 diabetes mellitus and type 2 diabetes mellitus with reports of the mean amplitude of glycemic excursion (MAGE), time in range (TIR), time above range (TAR), or time below range (TBR). Eligible trials were analyzed by fixed-effect model, random effect model, and meta-regression analysis accordingly. Results: In total, thirteen trials were included. Compared with the control group, physical activity intervention was significantly associated with increased TIR (WMDs, 4.17%; 95% CI, 1.11 to 7.23%, P<0.01), decreased MAGE (WMDs, -0.68 mmol/L; 95% CI, -1.01 to -0.36 mmol/L, P<0.01) and decreased TAR (WMDs, -3.54%; 95% CI, -5.21 to -1.88%, P<0.01) in patients with diabetes, but showed insignificant effects on TBR. Patients with higher baseline BMI levels was associated with a greater decrease in MAGE (b=-0.392, 95% CI: -0.710, -0.074), and patients with lower baseline HbA1c levels was associated with a greater increase in TBR during physical activities (b=-0.903, 95% CI: -1.550, -0.255). Conclusion: Physical activity was associated with significantly decreased glycemic variability in patients with diabetes. Patients with higher BMI might benefit more from physical activity therapy in terms of a lower MAGE. Hypoglycemia associated with physical activity treatment still warranted caution, especially in patients with intensive glycemic control.