Digital CBT was the only digital insomnia therapy to show a meaningful benefit in an analysis of 96 trials

The treatment that sleep specialists recommend first for chronic insomnia is a course of talking therapy, and it is hard to get. Cognitive behavioral therapy for insomnia takes several sessions with a trained clinician, and such clinicians are in short supply. A 2025 review put the problem in one line: the therapy is effective but limited by cost and therapist availability.

Software has moved into that gap. Sleep software now ranges from structured therapy courses to mindfulness apps, light-timing tools and virtual reality headsets, all described as digital therapeutics. Zihang Tong, Jingui Wang and colleagues at Tianjin University of Traditional Chinese Medicine set out to compare the main types against one another. Their analysis in BMC Medicine drew on 96 randomized trials, and one category came out clearly ahead. Digital cognitive behavioral therapy was the only approach whose effect on insomnia was large enough to matter to a patient.

What digital therapy for insomnia involves

Researchers define insomnia as difficulty initiating sleep, difficulty maintaining sleep, or early-morning awakening despite adequate opportunity for sleep, together with trouble functioning during the day. Cognitive behavioral therapy for it, usually shortened to CBT, is a package of techniques. People are taught to use the bed only for sleep, to limit the hours they spend in it so that sleep becomes more consolidated, to relax before bed and to question the anxious beliefs about sleeplessness that keep them awake.

Digital versions deliver the same program through a website or phone app over several weeks, with sleep diaries, lessons and automated feedback. Some include contact with a therapist and some run with no human involved.

The review sorted digital sleep products into six kinds. Alongside digital CBT, it covered a briefer behavioral version, mindfulness-based therapy, programs that adjust light and timing to shift the body clock, virtual reality relaxation, and neurofeedback delivered remotely, in which users try to change their own brain activity as shown on a screen.

How 96 trials of digital insomnia therapy were compared

The team searched four databases for randomized controlled trials in adults with sleep problems, whether or not they had a formal diagnosis. The search produced a total of 96 trials comprising 18,419 participants.

Few of those trials compared one digital therapy with another. Most compared a program with a waiting list, usual care, sleep education or a sham app. The authors therefore used a network meta-analysis, a method that links treatments through their shared comparison groups to estimate how each would fare against the others even where no head-to-head trial exists.

Most trials measured their results with the Insomnia Severity Index, a seven-question form on which people rate their sleep problems and how much those problems interfere with life. Higher scores mean worse insomnia. The reviewers also gathered sleep diary results, such as minutes taken to fall asleep, and the much scarcer measurements from sleep laboratories and wrist sensors.

They added one more test, and it shapes their conclusions. A difference can be statistically reliable and still too small for anyone to feel. For each outcome the reviewers set a minimum change that would count as meaningful to a patient. On the Insomnia Severity Index that minimum was 4 points.

Which digital insomnia therapies worked?

Digital CBT lowered insomnia scores by a little over 4 points more than the comparison conditions, on evidence the authors graded as moderately certain. It was the single result in the whole analysis to clear the meaningful-change bar.

On diary measures it helped, modestly. People using digital CBT fell asleep about 13 minutes sooner and spent about 17 fewer minutes awake during the night. Neither figure reached the reviewers’ thresholds of 20 and 30 minutes.

Mindfulness apps lowered insomnia scores by a little over 2 points, about half the effect of digital CBT and short of the 4-point bar. When the two were compared through the network, digital CBT did better by about 2 points.

For the other four categories there was little to report. Virtual reality reduced measured time awake at night in a handful of studies. Neurofeedback, brief behavioral therapy and body-clock programs showed no clear effect on any outcome, and each rested on few trials. The authors caution that the results derived from objective outcomes warrant cautious interpretation, and that most comparisons between one digital therapy and another remain insufficiently precise to rank them.

Two further comparisons bear on how digital CBT is used. Where trials had pitted digital CBT against the same therapy delivered face to face, the review found no clear difference. And within digital CBT, versions with a therapist in the loop did slightly better than fully automated ones.

How big is a 4-point change in insomnia severity?

Which yardstick to trust decides the answer. The 4-point threshold the reviewers adopted is a floor. When Charles Morin and colleagues at Université Laval validated the Insomnia Severity Index in 2011, they found that in a clinical sample it took a drop of about 8 points to be associated with moderate improvement as rated by an independent assessor. On that reading, the average benefit of digital CBT across these trials is noticeable but not transformative, about halfway to what a clinician would call a moderate improvement.

Averages hide a spread. Some people in these trials will have improved by far more than 4 points and some not at all, and the review could not say who is who.

Where the review sits beside guidelines and earlier trials

The result agrees with what guidelines already say about the therapy itself. The 2023 European Insomnia Guideline states that CBT for insomnia is recommended as the first-line treatment for chronic insomnia in adults of any age, either applied in-person or digitally. The American Academy of Sleep Medicine’s guideline is worded as an instruction to doctors: “We recommend that clinicians use multicomponent cognitive behavioral therapy for insomnia for the treatment of chronic insomnia disorder in adults”. The same guideline advises against relying on sleep hygiene tips alone.

Earlier pooled analyses had looked at digital CBT by itself. The 2025 review by Ji Woo Hwang and colleagues at Dongguk University in Seoul covered fully automated versions across 29 randomized controlled trials (RCTs) involving 9475 participants. It found moderate to large effects on insomnia severity, with the qualification that the automated version was less effective than therapist-assisted CBT-I, the abbreviation for CBT for insomnia.

The largest single trial points the same way. Colin Espie of the University of Oxford and colleagues evaluated 1711 participants with self-reported symptoms of insomnia, comparing a digital CBT program with sleep hygiene education. Daytime well-being and sleep-related quality of life improved more with the program, and the gains were still present at 24 weeks. Espie co-founded the company behind the program tested.

What the Tianjin analysis adds is the comparison across types. Earlier reviews could say that digital CBT works. They could not say whether a mindfulness app or a virtual reality headset would do as well, and on present evidence the answer is that they do not.

How reliable are the digital insomnia trials?

The reviewers rated every trial for risk of bias, meaning features of the design that could tilt the result, and the tally is sobering. Just two of the 96 were judged to be at low risk overall. Seventy-nine raised some concerns and 15 were at high risk.

One weakness comes with the territory. Nobody can be given a sleep app without knowing it, and the principal outcomes are questionnaires and diaries filled in by the same people. The authors write that participants are likely to be aware of their allocation, which may amplify expectancy effects and placebo responses. Someone who has spent six weeks on a program they hoped would help may rate their sleep more kindly than someone left on a waiting list.

Three other limits stand out. Results varied widely from trial to trial, more than the reviewers could explain. The analysis used only the measurements taken when treatment ended, so the authors say it is unable to address whether treatment effects wane or are sustained. And money is involved: they report that 40 of the 96 trials may have potential industry funding, introducing a risk of sponsorship bias.

What the review changes for people choosing a sleep program

It does not change guidelines, which already place CBT first. The authors themselves observe that current clinical guidelines and expert consensus remain relatively cautious in their endorsement of digital therapeutics as a class, and their own data suggest why: the label covers one approach with a substantial body of trials and five with very little.

Of the six kinds of digital sleep therapy tested, only CBT has shown a benefit large enough to notice, and that benefit is moderate, measured by self-report and unproven beyond the end of treatment.