Introduction: Internet-based interventions (IBI) can increase access to evidence-based treatments for mental disorders, but knowledge of their mechanisms of change is limited. Self-efficacy is an important common factor of psychotherapy and is especially interesting in IBI given its self-help focus. We investigated self-efficacy as an outcome, predictor/moderator, and mediator in IBI for mental disorders in randomized controlled trials.Methods: A systematic search was conducted across PsycINFO, PubMed, CINAHL, and Web of Science. Two reviewers selected studies, extracted data, and assessed bias. Effects were quantified using random effect models and supplemented by narrative syntheses and box score visualizations.Results: In total, 55 studies (N = 8798 participants) were included. IBI showed overall moderate effects on self-efficacy from pre- to post-treatment with guided interventions having the largest effect. Results were mixed regarding self-efficacy as a predictor/moderator, with three studies indicating that individuals with lower self-efficacy might benefit more from IBI. Self-efficacy emerged as a mediator through which IBI affected treatment outcomes.Conclusion: Overall, self-efficacy may be an influential factor in the efficacy of IBI for mental disorders and thus itself a valuable target of treatment. However, mixed results and methodological shortcomings of mediator studies call for further research, particularly concerning long-term impacts.