What is the public health significance of this article? Little is known about why cognitive behavioral therapy for insomnia leads to improvement. This study highlights the importance of reducing dysfunctional beliefs and, to some extent, monitoring and safety behaviors to achieve a positive treatment effect. Objective: To examine if the processes in the cognitive model mediate cognitive therapy (CT) and behavior therapy (BT) for insomnia. Method: Individuals diagnosed with insomnia disorder (n = 219) were randomized to telephone-supported internet-delivered CT (n = 72), BT (n = 73), or a wait-list (WL; n = 74). Cognitive processes (worry, dysfunctional beliefs, monitoring, and safety behaviors) proposed to maintain insomnia and treatment outcome (insomnia severity index) were assessed biweekly. Criteria for evaluating mediators were assessed via parallel process growth modeling and cross-lagged panel models. Results: Parallel process growth modeling showed that dysfunctional beliefs, monitoring, and safety behaviors significantly mediated the effects of both CT and BT. Cross-lagged panel models confirmed that dysfunctional beliefs and monitoring (approaching significance) influenced subsequent within-individual change in insomnia severity in CT. In BT, however, prior changes in insomnia severity predicted subsequent changes in worry and monitoring, and reciprocal influences among processes and outcomes were observed for dysfunctional beliefs and safety behaviors. Furthermore, the effect of safety behaviors on outcome was significantly larger for BT compared to CT. Conclusion: Together, the findings support the role of dysfunctional beliefs and monitoring as processes of change in CT and safety behaviors as a specific mediator in BT. Limited evidence was provided for worry as a mediator. The findings could improve clinical management and increase our conceptual understanding of insomnia and its maintaining factors by underscoring the relevance of these three processes for insomnia, as well as indicate important routes for future research, such as investigating how baseline presentations might moderate these mediations, for example moderated mediation.