Loneliness is a common phenomenon associated with several negative health outcomes. Current knowledge regarding interventions for reducing loneliness in randomised controlled trials (RCTs) is conflicting. The aim of the present work is to provide an overview of interventions to reduce loneliness, using an umbrella review of previously published systematic reviews and meta-analyses. We searched major databases from database inception to 31 March 2020 for RCTs comparing active versus non-active interventions for reducing loneliness. For each intervention, random-effects summary effect size and 95% confidence intervals (CIs) were calculated. For significant outcomes (p-value < 0.05), the GRADE (Grading of Recommendations Assessment, Development and Evaluation) tool was used, grading the evidence from very low to high. From 211 studies initially evaluated, seven meta-analyses for seven different types of interventions were included (median number of RCTs: 8; median number of participants: 600). Three interventions were statistically significant for reducing loneliness, that is, meditation/mindfulness, social cognitive training and social support. When applying GRADE criteria, meditation/mindfulness (mean difference, MD = -6.03; 95% CI: -9.33 to -2.73; very low strength of the evidence), social cognitive training (8 RCTs; SMD = -0.49; 95% CI: -0.84 to -0.13; very low strength of the evidence) and social support (9 RCTs; SMD = -0.13; 95% CI: -0.25 to -0.01; low strength of the evidence) significantly decreased the perception of loneliness. In conclusion, three intervention types may be utilised for reducing loneliness, but they are supported by a low/very low certainty of evidence indicating the need for future large-scale RCTs to further investigate the efficacy of interventions for reducing loneliness.