Objectives: This article reviews the effects of physical activity (PA) interventions on social isolation, loneliness or low social support in older adults. Design: Systematic review and meta-analysis of randomised controlled trials (RCTs). Method: MEDLINE, EMBASE, PsycINFO, the Cochrane CENTRAL, CINAHL, were screened up to February 2017. RCTs comparing PA versus non-PA interventions or control (sedentary) condition were included. Risk of bias was assessed using the 12 criteria Cochrane Review Book Group risk of bias. The outcome measures were: social isolation, loneliness, social support, social networks, and social functioning. Standardised mean differences (SMDs) with associated 95% confidence intervals (CIs) were calculated for continuous outcomes. Meta-analysis was performed using a random effects model. Results: The search strategy identified 38 RCTs, with a total of 5288 participants, of which 26 had a low risk of bias and 12 had a high risk of bias. Meta-analysis was performed on 23 RCTs. A small significant positive effect favouring the experimental condition was found for social functioning (SMD=0.30; 95% CI, 0.12 to 0.49; P=0.001) with strongest effects obtained for PA interventions, diseased populations, group exercise setting, and delivery by a medical healthcare provider. No effect of PA was found for loneliness, social support, or social networks. Conclusion: This review shows, for social functioning, the specific aspects of PA interventions can successfully influence social health. PA did not appear to be effective for loneliness, social support and social networks.