This review evaluates the evidence on what interventions are effective in reducing public stigma towards people with severe mental illness, defined as schizophrenia, psychosis or bipolar disorder. We included 62 randomised controlled trials of contact interventions, educational interventions, mixed contact and education, family psychoeducation programs, and hallucination simulations. Contact interventions led to small-to-medium reductions in stigmatising attitudes (d = 0.39, 95% CI: 0.22 to 0.55) and desire for social distance (d = 0.59, 95% CI: 0.37 to 0.80) post-intervention, but these were reduced after adjusting for publication bias (d = 0.24 and d = 0.40, respectively). Effects did not vary by type or length of contact. Effects at follow-up were smaller and not significant. Education interventions led to small-to-medium reductions in stigmatising attitudes (d = 0.30, 95% CI: 0.14 to 0.47) and desire for social distance (d = 0.27, 95% CI: 0.08 to 0.46) post-intervention. Small improvements in social distance persisted up to 6 months later (d = 0.27, 95% CI: 0.05 to 0.49), but not attitudes (d = 0.03, 95% CI: -0.12 to 0.18). The combination of contact and education showed similar effects to those that presented either intervention alone, and head-to-head comparisons did not show a clear advantage for either kind of intervention. Family psychoeducation programs showed reductions in stigma post-intervention (d = 0.41, 95% CI: 0.11 to 0.70). The effectiveness of hallucination simulations was mixed. In conclusion, contact interventions and educational interventions have small-to-medium immediate effects upon stigma, but further research is required to investigate how to sustain benefits in the longer-term, and to understand the active ingredients of interventions to maximise their effectiveness.