Acupuncture as one of the alternative therapies for insomnia is widely used in Asia and increasingly employed in western countries. To provide updated evidence from randomized controlled trials (RCTs) on the effectiveness and safety of acupuncture for primary insomnia. A comprehensive literature search in 11 databases was conducted from January 2008 to October 2017. Two authors independently extracted data and assessed risk of bias independently. Statistical analysis was performed using RevMan 5.3 software. According to predefined protocol, we combined data in meta-analysis and performed trial sequential analysis when appropriate. Grading of Recommendations Assessment, Development, and Evaluation was also conducted to assess the quality of evidence. A total of 73 RCTs involving 5533 participants were analyzed. The pooled results showed better effect from real acupuncture than no treatment (mean difference [MD] -5.58, 95% confidence interval [CI] -6.85 to -4.31,  = 0%,  < 0.00001, 2 trials, fixed effect model, 105 participants) on reducing Pittsburgh Sleep Quality Index (PSQI) scores with "very low quality" evidence. Acupuncture plus drugs showed better improvement than drugs alone on decreasing the PSQI total scores (MD -3.17, 95% CI -4.74 to -1.61,  = 72%, 4 trials, random-effects model (REM),  < 0.0001, 253 participants, low quality). Similar benefit favored acupuncture compared with no treatment (MD -8.46, 95% CI -9.59 to -7.33,  = 0%,  < 0.00001, 2 trials, 65 participants). Acupuncture showed more benefit than estazolam on PSQI (with enough statistical power). Athens Insomnia Scale (MD -1.64, 95% CI -2.40 to -0.89,  = 0%,  < 0.0001, 3 trials, fixed-effects model, 180 participants) or SPIEGEL (MD -2.86, 95% CI -3.54 to -2.18,  < 0.00001,  = 0%, 5 trials, fixed-effects model, 326 participants) with "very low-quality" evidence. Furthermore, low-quality evidence showed less adverse events from acupuncture than western medications (risk ratio 0.23, 95% CI 0.11-0.48,  = 56%,  < 0.0001, 11 trials, REM, 914 participants). Publication bias was likely present based on the PSQI total scores. The summary estimates indicate that acupuncture might result in improvement than no treatment on PSQI scores and appears safe. However, the quality of the evidence is varied from very low to low due to the potential risk of bias and inconsistency among included trials. Further large sample size and rigorously designed RCTs are still needed.