Why the comparison group changes the size of the benefit
An effect size is a measure of how far apart two groups end up, so it is only as meaningful as the group on the other side. A waitlist control receives nothing, so any improvement from the digital program stands out clearly. A sleep education or usual care control gives people something to do — reading material, general advice, routine care — and some of that can move sleep symptoms on its own, which shrinks the gap the program has to clear.
The 2025 meta-analysis tested this directly: across 29 randomized controlled trials covering 9,475 participants, the type of control group moderated the outcomes, meaning the same intervention looked more or less impressive depending on what it was measured against [1]. The authors describe the overall effects on insomnia severity as moderate to large, and note that the program had a significant impact when contrasted with most control groups [1]. The word "most" is doing real work there — it signals that the advantage was not uniform across every comparison.
What the trials actually compared against
The most common comparison in this literature is not a waitlist at all. Several of the trials behind these findings used an education-style control: a nine-week patient education program in a pilot study of heavy drinkers with insomnia [3], and an online education control in a twenty-year meta-analysis of the SHUTi program [5]. A Norwegian trial protocol likewise plans to compare fully automated digital CBT-I against a static online information webpage in people with multiple sclerosis [2].
That matters for interpreting the headline numbers. When the control is educational material rather than nothing, the digital program is being credited only for the advantage it holds over credible information — a stiffer test than a waitlist. So a moderate-to-large effect against education controls is a more conservative signal than the same number against a waitlist would be, even though the meta-analysis did not report the two side by side [1][5].
Finishing the program is not the same as using it
One of the more useful findings concerns what did not explain the variation in results. The meta-analysis examined completion rates and found they did not moderate outcomes, while control group type did [1]. In other words, how many people made it to the end of the program told the researchers little about how well it worked; what it was compared against told them more.
The authors read this as a hint that engagement — actually doing the exercises, tracking sleep, adjusting behavior — matters more than simply reaching the final module [1]. That is an interpretation rather than a demonstrated mechanism, but it fits the pattern: a program can be finished passively, and a finished-but-passive program would not be expected to shift insomnia severity much.
Where the fully automated version stands against a therapist
The same meta-analysis found that fully automated digital CBT-I was less effective than CBT-I delivered with a therapist's involvement [1]. The authors conclude that a hybrid model combining therapist support would be more beneficial, while still describing the fully automated version as a promising option for managing insomnia given the cost and availability limits of therapist-delivered care [1].
Some trials also point to effects beyond sleep itself. In heavy drinkers with insomnia, the SHUTi program improved both sleep and drinking outcomes relative to a patient education control, and the effect on how much people drank — though not how often — appeared to run through the improvement in sleep [3]. That is a preliminary finding from a small pilot, and it concerns a specific high-risk group, so it should not be generalized to general insomnia populations. A related analysis of the same program reported consistently larger within-group reductions on some outcomes for one sex than the other, but the comparison group there was an active control rather than a waitlist or usual care [4].
Sources used in this answer
Systematic review and meta-analysis on fully automated digital cognitive behavioral therapy for insomnia
Across 29 randomized trials (9,475 participants), fully automated digital CBT-I showed moderate-to-large effects on insomnia severity versus controls, control group type moderated outcomes, completion rate did not, and it was less effective than therapist-assisted CBT-I.
Digital cognitive behavioral therapy for insomnia compared with digital patient education about insomnia in people with multiple sclerosis in Norway: study protocol for …
A study protocol plans to compare fully automated digital CBT-I with a static online webpage control in people with multiple sclerosis, citing small to moderate effect sizes for dCBT-I.
0320 Improvements in Sleep & Drinking Outcomes Among Heavy Drinkers Following a Digital CBT-I Program
In a pilot of 53 heavy drinkers with insomnia, the SHUTi digital CBT-I program improved sleep and drinking outcomes relative to a nine-week patient education control, with effects on drinking quantity but not frequency mediated by sleep improvements.
Sex differences in sleep and alcohol consumption outcomes following a digital insomnia intervention
Within-group effect sizes following the SHUTi digital CBT-I program differed by sex for sleep and alcohol outcomes, in a comparison against an active control group.
Sleep Healthy Using the Internet: A Meta-Analysis of Twenty Years of Research on a Digital Cognitive Behavioral Therapy for Insomnia Program
A twenty-year meta-analysis of the SHUTi digital CBT-I program, using an online education control, reports effect sizes demonstrating efficacy.