The majority of people with dementia (PwD) live in the community. Compared to institutionalised PwD, their access to formalised music therapy is limited. Initial works suggest that non-therapist-led music-based interventions (MBIs) may be an accessible and effective alternative. The aim of this review was, therefore, to synthesise evidence on MBIs for community-dwelling PwD. We systematically searched electronic databases (PubMed, PsycInfo, Web of Science) for records reporting on controlled studies of MBIs delivered to community-dwelling PwD. Two reviewers independently screened records according to inclusion/exclusion criteria. A total of 15 relevant publications reporting on 14 studies were initially identified and assessed using the Cochrane risk-of-bias tool for randomised trials (RoB 2) and the risk of bias. In non-randomised studies of interventions (ROBINS-I) tool. A total of 11 records of 10 studies, with a total of n = 327 PwD, were included in the synthesis. MBIs consisted either of singing or music listening interventions and were variable in duration. MBIs had immediate positive effects on cognition. Short-term MBIs (lasting 1-4 months) benefited cognition, anxiety and pain. Evidence on depressive symptoms was conflicting. The benefits of longer term MBIs (lasting 6+ months) were less apparent. According to GRADE criteria, the overall quality of evidence was moderate to low. The inconsistency in designs, procedures and measures prevents specific conclusions at this stage. Still, the diversity observed in existing studies suggests that there are multiple interesting avenues for researchers to pursue, including the involvement of informal caregivers in MBI delivery. Future studies need to ensure adequate reporting to facilitate continued development. The protocol of this review was pre-registered with the Prospective Register of Systematic Reviews (PROSPERO, Registration Number: CRD42020191606).