Abstract Background Virtual reality (VR) may improve psychological treatments for psychotic disorders. We investigated the effects of VR-based cognitive behavior therapy for paranoid ideation (VR-CBTp) compared to standard CBTp. Methods We conducted a pragmatic, single-blind, randomized clinical trial in seven mental health centers across the Netherlands and Belgium. A total of 98 participants with a psychotic spectrum disorder and paranoid ideation were randomized to a maximum of 16 sessions of VR-CBTp (n = 48) or CBTp (n = 50). The primary outcome was momentary paranoia, measured with the experience sampling method (ESM) at posttreatment. Secondary measures, assessed at baseline, posttreatment, and 3-month follow-up, included symptoms (paranoia, hallucination, depression, cognition, and anxiety related), social functioning, self-esteem, and schemes. Results Both groups showed reductions in momentary paranoia between baseline and posttreatment (n = 56, b = −15.0, effect size [ES] = 0.65), but those were greater for VR-CBT (interaction b = 8.3, ES = 0.62). Reductions remained at follow-up (n = 50, b = −10.7, ES = 0.57) but not the interaction. Limited ESM compliance resulted in data loss; however, secondary paranoia measures did confirm improvements (ES range = 0.66–1.15, n = 78–81), but not the interaction. Both groups improved in symptoms, self-esteem, and social functioning. Interaction effects in favor of VR-CBTp were found for safety behavior, depression, and self-esteem at posttreatment, and self-esteem and anxiety at follow-up. For VR-CBTp, 37.5% did not complete treatment; for CBTp, this was 24.0%. Completers, on average, received 12.7 (VR-CBTp: standard deviation [SD] = 3.9) and 15.1 (CBTp: SD = 2.5) sessions. Conclusions Both CBTp and VR-CBTp are efficacious treatments for paranoid ideation, but VR-CBTp may be somewhat more effective. Limitations concern missing primary outcome data and a lower sample size than anticipated.