The mental health app features people use most are not the ones that work

Most mental health apps lead with the same handful of features — psychoeducation, relaxation exercises, mindfulness, mood tracking — and a meta-regression of 169 trials finds those are not the ingredients most strongly tied to getting better [s1]. The techniques that best predicted a drop in depression were desensitisation, stimulus control and activity scheduling; for anxiety it was exposure [s1]. That gap, between what apps offer and what works, is the most actionable finding in the field.

What was analysed

Researchers coded the contents of mental health apps against 34 pre-registered cognitive-behavioural elements and then modelled the association between each element and symptom change [s1]. The pooled dataset was large: 169 trials with 1,137 measurement timepoints, 41,807 participants, a mean age of 34.3 years, and 72.9% female [s1]. Rather than ask the usual question — does app therapy beat a waitlist — the analysis asked a more useful one: given that these products are bundles of techniques, which techniques are carrying the effect.

The mismatch

The most frequently included ingredients were psychoeducation, relaxation, mindfulness and self-monitoring [s1]. Many of the coded elements were only weakly to moderately effective, and the strength varied considerably from one to the next [s1].

The elements most strongly and robustly associated with improvements in depression were desensitisation, stimulus control and activity scheduling [s1]. For anxiety, the exposure-based elements did the most work [s1]. These are the more demanding, more clinically specific parts of cognitive behavioural therapy — the components that ask a user to schedule activity against low mood, to confront rather than avoid a feared situation, to restructure the cues around a behaviour. They are harder to build into a friendly app and harder to get people to complete, which may be exactly why the gentler techniques dominate the market.

Two elements stood out for doing nothing: graded tasks and personal-strengths exercises were classed as ineffective [s1]. And more was more — interventions that incorporated a greater number of active elements were more effective overall [s1]. A single-technique app, in other words, is starting from behind.

It fits the earlier trial data

This is not an isolated result. The largest efficacy meta-analysis of app-supported interventions, across 66 randomised trials, had already found that the apps producing the larger effects were the ones built on cognitive behavioural therapy and paired with professional guidance and reminders to engage [s2]. The meta-regression explains part of why: CBT is where the high-yield ingredients live.

The case of mindfulness is instructive because it is both ubiquitous and only modestly effective. A dedicated meta-analysis of 34 randomised trials of mindfulness-meditation apps (7,566 participants) found real but small post-test effects — perceived stress g = 0.46, depression g = 0.33, anxiety g = 0.28, psychological well-being g = 0.29 — and no significant effect at all on general distress (g = 0.10) or general well-being (g = 0.14) [s3]. Mindfulness is not useless; it is simply not the strongest tool in the box, despite being one of the most common [s1][s3].

What it means for a reader

The practical read is that the label matters less than the mechanism. "Mental wellness," "calm" and "mood" describe a category, not a method. On the current evidence, an app is more likely to help if it delivers structured cognitive behavioural techniques — activity scheduling, exposure, stimulus control — and if it includes several of them rather than one [s1]. The apps that do this well tend to be the CBT-based, guided programmes that also performed best in the head-to-head trial literature [s2], several of which — for insomnia and panic disorder — have been tested far more rigorously than the average product in an app store.

None of this tells a reader which specific app to use, and the meta-regression is an observational map of associations, not a randomised test of any one feature [s1]. But it reframes the choice usefully: the question is not whether an app is soothing to open, but whether it contains the parts of therapy that the evidence says actually change symptoms.

Sources

• [s1] Identifying what works in mental health apps through meta-regression analyses of 169 trials — npj Digital Medicine (2026)
• [s2] The efficacy of app-supported smartphone interventions for mental health problems: a meta-analysis of randomized controlled trials — World Psychiatry (2019)
• [s3] The efficacy of mindfulness meditation apps in enhancing users' well-being and mental health related outcomes: a meta-analysis of randomized controlled trials — Journal of Affective Disorders (2020)