We performed a systematic review and Bayesian network meta-analysis to determine atypical antipsychotics that are effective and safe for the treatment of psychosis in Parkinson’s disease (PD). We conducted a comprehensive literature search using PubMed/MEDLINE, Cochrane Library, and Japana Centra Revuo Medicina (Ichu-shi Web). We used randomized controlled trials evaluating the utility of atypical antipsychotics for the treatment of psychosis in PD using the Brief Psychiatric Rating Scale (BPRS) and the Unified PD rating Scale parts III (UPDRS-III) as the endpoints. Posterior distributions of mean differences between each treatment and placebo were estimated using Bayesian network meta-analysis. The distributions describing each treatment effect were expressed as means (95% credible intervals). Ten trials involving any two treatment arms using clozapine (64 subjects in four trials), olanzapine (99 subjects in three trials), quetiapine (79 subjects in five trials), risperidone (five subjects in one trial), or placebo (156 subjects in seven trials) were finally included in the present study. Pooled estimates of each posterior distribution based on the BPRS were as follows: clozapine, −2.0 (−6.7 to 2.7); olanzapine, 0.5 (−2.3 to 3.4); quetiapine, 0.3 (−3.9 to 4.5); and risperidone, −4.7 (−57.4 to 53.3). Based on the UPDRS-III, the pooled estimates were clozapine, 0.7 (−3.8 to 4.3); olanzapine, 2.8 (0.8 to 5.1); quetiapine, 3.3 (−0.7 to 5.8); and risperidone, 4.5 (−57.7 to 63.4). Although clozapine had an effective and relatively safe profile, all atypical antipsychotics included in the present study may be unsafe, as they may worsen motor function when compared to placebo.