This meta-analysis mainly examined the effect size of exposure and response prevention (ERP) for obsessive-compulsive disorder (OCD) when compared with different control conditions, and the difference in the efficacy of different variants of ERP in the treatment of OCD. Thirty studies were included, involving 39 randomized controlled trials with 1793 participants, from 30 studies up to January 18, 2022. Hedge's g was calculated using random-effects models. The results showed that ERP had a definite effect on OCD (g = 0.37), and its effect was significant when the control condition was placebo (g = 0.97) or drug (g = 0.59). However, ERP did not show statistical differences with other therapies in improving OCD (g = -0.07). In addition, we found that both therapist and self-controlled exposure (at the same time as the therapist controls, self-control is exercised after the therapy session) and total response prevention can better improve OCD symptoms. In addition, compared with the control group, ERP reduced depression (g = 0.15) and anxiety symptoms (g = 0.23) in patients with OCD. Meta-regression results showed that the longer the length of sessions, the better the treatment effect (t = 2.41, p = 0.022).