Background: Tardive dyskinesia (TD) is characterized by abnormal and involuntary movements. Importantly, TD could cause considerable personal suffering and social and physical disabilities. Aims: This meta-analysis based on randomized controlled trials (RCTs) systematically assessed the therapeutic effect and tolerability of melatonin for TD in schizophrenia. Methods: A computerized and systematical search of both Chinese (Wanfang Data, Chinese National Knowledge Infrastructure (CNKI), SINOMED) and English (PubMed, PsycINFO, Embase, Cochrane Library databases) databases, from their inception until June 8, 2017, was conducted by two independent authors. The severity of TD symptoms were the primary outcome measure and analyzed using a random effects model by the Review Manager (RevMan) Version 5.3. Quality evaluation of included RCTs was conducted using the Cochrane risk of bias and Jadad scale. The GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) system recommendation grading method was used to assess the overall quality level of meta-analytic outcomes. Results: Four RCTs (n=130) were identified and analyzed. Three RCTs used double blind and 1 RCT used masked assessors using the Cochrane risk of bias, and 3 RCTs were rated as high quality based on Jadad scale. Compared with the control group, adjunctive melatonin was superior in reducing the severity of TD as measured by the Abnormal Involuntary Movement Scale (AIMS) (4 RCTs, n=130, weighted mean difference (WMD): -1.52 (95% confidence intervals (CI): -3.24, 0.20), p=0.08; I2=0%) although the improvement did not reach a significant level. The overall evidence quality of the improvement of TD symptoms, according to GRADE approach, was rated as "Low". The data on the ADRs and cognitive effect were limited. Conclusions: This meta-analysis shows that melatonin has potential for improving TD symptoms in schizophrenia. Future higher quality and larger RCTs are warranted to confirm the findings.