How effective is fully automated digital CBT-I compared with therapist-assisted CBT-I?

How well does fully automated digital CBT-I work on its own?

Fully automated digital CBT-I—a web or app program that delivers cognitive behavioral therapy for insomnia without any human therapist involvement—does help people sleep better. Across 29 randomized controlled trials with 9,475 participants, it produced moderate to large improvements in insomnia severity compared with control groups [1]. A separate meta-analysis of 22 trials found that digital CBT-I, including fully automated versions, had a large effect on sleep outcomes and also reduced depression and anxiety symptoms by small to moderate amounts [5].

The benefits are not just statistical. In one randomized trial, 46% of people who used fully automated digital CBT-I achieved remission from insomnia, compared with 22% of those who received online sleep education [3]. That means roughly half of users still had clinically significant insomnia after the program—a useful benchmark for what fully automated CBT-I can and cannot do.

Why does adding a therapist make CBT-I work better?

The gap between automated and therapist-assisted CBT-I appears in direct comparisons. In a secondary-care head-to-head study, fully automated digital CBT-I was inferior to individual CBT-I: at 9 weeks, 43% of automated users responded versus 70% of individual CBT-I users, and remission rates were 18% versus 52%. At about 33 weeks, remission was 24% versus 56% [2]. The automated program worked, but it did not work as well as having a person guide the treatment.

One likely reason is adherence—not just finishing the program, but actually doing the behavioral work it requires. The large meta-analysis found that completion rate did not explain differences in outcomes, but treatment adherence did, suggesting that how consistently people apply the techniques matters more than whether they simply finish the modules [1]. A therapist may help people stay engaged with the harder parts, such as sleep restriction, and adjust the plan when problems arise.

When does the automated-versus-therapist gap matter most?

The difference is largest for people who do not remit with automated treatment. In a stepped-care trial, participants first received fully automated digital CBT-I; those who did not achieve insomnia remission were then randomized to face-to-face CBT-I or a control. Face-to-face CBT-I produced a 70% remission rate in this group, compared with 23% in the control, and those who received face-to-face CBT-I had a much lower rate of new depression (5%) than the control group (15%) [3]. This suggests that automated CBT-I can be a reasonable first step, with therapist-assisted CBT-I reserved for those who need more.

For people with comorbid depression or anxiety, therapist-delivered CBT-I via videoconference achieved 66% response and 43% remission in a real-world cohort, and two trials found it non-inferior to in-person CBT-I [2]. Fully automated digital CBT-I in comorbid samples still improved insomnia and reduced depression and anxiety symptoms, but the same review noted that fully automated versions may underperform individual CBT-I, supporting guided or stepped-care models [2].

Does some human support help even if it is not full therapy?

The evidence on partial human support—such as coaching alongside a digital program—is less clear from these studies. One randomized trial directly compared digital CBT-I with different types of coaching, but the available abstract does not report the comparative outcomes [4]. What the larger body of evidence does show is a consistent pattern: fully automated digital CBT-I is a useful, accessible option that helps many people, but therapist-assisted CBT-I tends to produce higher response and remission rates, particularly for those who do not improve on their own [1][2][3].

Sources used in this answer

Related research questions