BACKGROUND: Robot-assisted surgery (RAS) has potential panspecialty surgical benefits. High-quality evidence for widespread implementation is lacking. This systematic review aimed to assess the RAS evidence base for the quality of randomized evidence on safety and effectiveness, specialty 'clustering', and outcomes for RAS research. METHODS: A systematic review was undertaken according to PRISMA guidelines. All pathologies and procedures utilizing RAS were included. Studies were limited to RCTs, the English language and publication within the last decade. The main outcomes selected for the review design were safety and efficacy, and study purpose. Secondary outcomes were study characteristics, funding and governance. RESULTS: Searches identified 7142 titles, from which 183 RCTs were identified for data extraction. The commonest specialty was urology (35·0 per cent). There were just 76 unique study populations, indicating significant overlap of publications; 103 principal studies were assessed further. Only 64·1 per cent of studies reported a primary outcome measure, with 29·1 per cent matching their registration/protocol. Safety was assessed in 68·9 per cent of trials; operative complications were the commonest measure. Forty-eight per cent of trials reported no significant difference in safety between RAS and comparator, and 11 per cent reported RAS to be superior. Efficacy or effectiveness was assessed in 80·6 per cent of trials; 43 per cent of trials showed no difference between RAS and comparator, and 24 per cent reported that RAS was superior. Funding was declared in 47·6 per cent of trials. CONCLUSION: The evidence base for RAS is of limited quality and variable transparency in reporting. No patterns of harm to patients were identified. RAS has potential to be beneficial, but requires continued high-quality evaluation.