As a popular method, virtual reality (VR) is still controversial in its effect on the balance function of patients with Parkinson’s disease. This systematic review aims to discuss such effects of VR and to compare it with that resulting from traditional therapies. A comprehensive search was conducted for randomized controlled trials published from 2000 to 2020 through the following databases: PubMed, Web of Science, CINAHL, Embase, Cochrane Library. Fifteen articles were included for the systematic review. An evaluation on their methodological qualities was performed using the PEDro scale, followed by an assessment of their risk of biases in accordance with the Cochrane Handbook for Systematic Reviews of Interventions for quality assessment. In terms of dynamic balance, the BBS score of the VR group was significantly improved when compared with the control group (SMD = 0.52, 95% CI = 0.31–0.73). However, no significant difference was observed between the two groups on TUG (SMD = −0.26; 95% CI = −0.62–0.1; p = 0.16). Besides, the VR group also showed better results in improving patients’ static balance, balance confidence, and quality of life. A funnel plot was created to investigate the effects of each study included in the meta-analysis in order to identify any existing publication bias. This systematic review shows that the application of VR leads to more significant improvement in the balance of patients with Parkinson’s disease than having them perform traditional exercises. It can be used as an auxiliary method of rehabilitation.