Objective This is a meta‐analysis of randomized controlled trials (RCTs) of the efficacy of melatonin in Alzheimer's disease (AD). Methods Both English (PubMed, PsycINFO, Embase, Cochrane Library databases, and the Cochrane Controlled Trials Register) and Chinese (WanFang Database, Chinese Biomedical Database, and China Journal Net) databases were systematically and independently searched by two authors from their inception until 1 March 2016. Weighted and standard mean differences (SMDs), risk ratio (RR) ±95% confidence intervals (CIs) were calculated. In all cases, the random effects model was used. Results Seven studies (n = 462) with the duration ranging from 10 days to 24 weeks were identified and analyzed. AD patients receiving melatonin treatment showed prolonged total sleep time at night (n = 305; SMD: 0.26, 95% CI: 0.01 to 0.51, I2 = 9%, p = 0.04). Melatonin did not improve cognitive abilities assessed by the mini‐mental state examination and the Alzheimer's Disease Assessment Cognitive Subscale. The discontinuation rate was similar between the melatonin and placebo groups (n = 453, RR = 0.77, 95% CI: 0.51 to 1.16, I2 = 0%, p = 0.21). Conclusion Melatonin appears to be effective and safe in improving sleep quality in patients with AD. Copyright © 2016 John Wiley & Sons, Ltd.