Objectives We performed a systematic review and meta-analysis to assess the effects of physical activity in preventing gestational diabetes mellitus (GDM). Search strategy We searched the literature in six electronic databases and bibliographies of relevant articles. Selection criteria We included randomised controlled trials on pregnant women who did not have GDM and other complications previously and had increased physical activity as the only intervention. The risk of developing GDM was documented separately for the intervention and control groups. Data collection and analysis Two reviewers extracted data and assessed quality independently. Data from the included trials were combined using a fixed-effects model. The effect size was expressed as relative risk (RR) and 95% CI. Main results Of the 1110 studies identified, six randomised controlled trials met the inclusion criteria. In three trials, the incidence of GDM was lower in the intervention group than in the control group, whereas two trials showed a higher incidence of GDM in the intervention group and the remaining trial found no GDM in either the intervention or control group. The meta-analysis resulted in a relative risk (RR) of GDM of 0.91 (95% CI 0.57 to 1.44), suggesting no significant difference in the risk of developing GDM between the intervention and the control groups. No indication of publication bias was found. Conclusions Evidence was insufficient to suggest that physical activity during pregnancy might be effective to lower the risk of developing GDM.