Mental health apps are everywhere. The app stores list more than ten thousand of them, marketed for everything from sleep to grief to schizophrenia. The research base behind that catalogue is much smaller, and a careful read of it produces a less dramatic picture than the marketing copy suggests. Apps work, on average, a bit better than nothing, and quite a lot less well than a competent therapist. The honest article on whether mental health apps work is mostly an article about effect sizes, comparators, and adherence.
The size of the gap between marketing and evidence
Start with the gap. Larsen and colleagues went through the science-flavoured claims that mental health apps make in the app stores and checked them against published evidence.6 Most claims were vague ("clinically validated," "evidence-based"), referred to general scientific principles rather than the specific product, or pointed to studies that had not actually been done on the app being sold. A small minority of apps had any randomised trial data at all. That is the baseline. When someone asks whether mental health apps work, the answer depends almost entirely on which app, used how, by whom.
The other baseline is that "mental health app" is a category as broad as "book." A meditation timer, a CBT thought-record journal, an AI chatbot, a symptom tracker for bipolar disorder, and a guided-self-help program for postnatal depression are all "mental health apps." Aggregate effect sizes across that category are useful as a sanity check on the marketing, not as a guide to what any specific tool will do for any specific person.
What the meta-analyses actually find
The single most-cited number in this field comes from Linardon and colleagues' 2019 meta-analysis in World Psychiatry.1 Sixty-six randomised controlled trials, around 7,400 participants, app-based interventions versus various control conditions. Pooled effects were small-to-moderate across depression, generalised anxiety, social anxiety, suicidality, stress, and quality of life, with Hedges' g in the 0.20–0.45 range depending on outcome. Two findings inside that paper matter more than the headline. Effects were larger when the app used CBT-style components, and larger when the app was used as a complement to in-person care rather than as a stand-alone treatment.
Firth et al. 2017 looked specifically at depression in 18 trials with around 3,400 participants and found a standardised mean difference around 0.38 versus inactive control conditions, shrinking against active comparators.3 Their parallel anxiety meta-analysis, nine trials, came in at roughly 0.33 with the same pattern: a real but modest effect, larger when the comparator did nothing.4
Goldberg et al. 2022 went one level higher and pooled 14 meta-analyses of mobile-phone interventions across mental-health outcomes.2 The conclusion of that meta-review is the one to remember. Smartphone interventions reliably outperform inactive control groups (waitlist, no treatment) but rarely beat active comparators (a generic wellness app, in-person therapy, structured psychoeducation). The signal is real. It is not large.
What this means in plain language
Apps work better than nothing. They roughly match other low-intensity interventions like bibliotherapy or supported self-help. They do not replace therapy for clinical-level conditions. Effect sizes are clinically modest, not magic.
The other half of the picture is adherence. Trial-level dropout looks reasonable because participants are screened, paid, and reminded. Real-world adherence is grim. Across multiple analyses, 70–90% of people who download a mental health app stop using it within two weeks. An app that produces a g of 0.4 in a 12-week trial does very little for someone who uses it for nine days. The number that matters in real life is not the trial effect size; it is the trial effect size multiplied by whether you actually open the app.
What separates the apps that work from those that don't
The same three predictors keep showing up. Apps with CBT-derived components (thought records, behavioural activation prompts, structured exposure work, scheduled mood and emotion entries) produce stronger effects than apps built around generic wellness content.12 This is the cleanest signal in the literature.
The second predictor is human contact, even minimal. Across both Linardon 2019 and the Goldberg meta-review, apps with any form of asynchronous human guidance — a coach, a therapist checking notes once a week, a peer-support layer — produced effects roughly twice the size of fully self-guided apps. "Guided self-help" is the historical term, and it earns its name. A small amount of accountability changes outcomes more than most feature additions.
The third predictor is consistent use. This sounds obvious until you notice how much app design optimises for first-week novelty rather than week-six retention. Streaks help some users and demoralise others. Notifications produce diminishing returns and outright fatigue. The apps that show up in the meta-analyses with positive effects tend to be the ones people open repeatedly without performative encouragement.
Where the field is going
The 2021 World Psychiatry review by Torous and colleagues maps the next wave: AI chatbots, passive sensing from phone sensors, integrated platforms that coordinate with clinicians, virtual reality exposure.5 Each of these has early supporting data and considerably more enthusiasm than data. Chatbot studies are small and short. Mood prediction from passive sensing is technically interesting and clinically unproven at the individual level. Integrated platforms work in the trials they run and have not yet been tested at scale.
The reasonable read of the Torous review is that the directions are sensible and the evidence is not yet there. A consumer choosing an app today should base the choice on the meta-analytic ground, which is plain CBT-style apps used consistently, with light human contact if available.
Reasonable expectations for users
For mild-to-moderate symptoms, an app with CBT components, used a few times a week, will probably help a little. The honest comparator is "more than journaling on paper, less than therapy." That is a useful slot to fill for self-awareness, between-session homework, and tracking patterns over weeks.
For severe depression, active suicidality, PTSD, OCD, or psychosis, an app is not an appropriate primary treatment. The meta-analyses do not show effects of that size, and the trials largely excluded those populations. The right move there is a clinician.
Colors covers the components the meta-analyses identify as load-bearing — granular emotion labelling, CBT-style thought records, structured between-session journaling — and stays out of the territory the data are weakest on, namely chatbot "AI therapists" and mood prediction. The honest claim: a useful adjunct, not a treatment in itself.
Frequently asked questions
Do mental health apps actually work?
On average, yes — modestly. The largest meta-analysis to date, Linardon et al. 2019 in World Psychiatry, pooled 66 randomised controlled trials with around 7,400 participants and found small-to-moderate effects on depression, anxiety, stress, suicidality, and quality of life. The effects were largest when apps used CBT-style techniques and when the app supported, rather than replaced, contact with a clinician. They were smaller against active controls than against waitlist, and adherence outside trials is much lower than inside them.
How big is the gap between marketing claims and evidence?
Large. There are over 10,000 apps in the app stores marketed for mental health, and only a small fraction have any randomised trial data behind them. A 2019 evaluation in npj Digital Medicine reviewed the scientific claims made by mental health apps in the Google Play and App Store and found most claims were either vague, unsupported by published evidence, or referred to general scientific principles rather than the specific app.
Are apps a replacement for therapy?
No. Across multiple meta-analyses, smartphone interventions match other low-intensity interventions and beat doing nothing, but they do not outperform face-to-face therapy for clinical-level depression, anxiety, or PTSD. The reasonable framing is adjunct, not substitute — between-session homework, journaling, mood tracking, and psychoeducation, with a clinician in the loop for anything serious.
What features separate apps that work from apps that do not?
Three things, fairly consistently across meta-analyses. First, CBT-derived components — thought records, behavioural activation, exposure prompts — predict larger effects than non-specific wellness content. Second, even brief asynchronous human contact roughly doubles effect sizes versus fully self-guided use. Third, sustained use beats clever one-time features; most users abandon apps within two weeks, and the apps that retain attention tend to also retain effect.
What about AI chatbots and mood prediction?
Early evidence, much smaller than the evidence base for app-delivered CBT. The 2021 Torous et al. review in World Psychiatry treats chatbots, passive sensing, and integrated platforms as promising directions where the data has not yet caught up to the marketing. Treat strong claims here with caution; the meta-analytic ground is in plain CBT-style apps used consistently.
Not medical advice
This article is for informational and educational purposes only. It does not constitute medical advice and should not replace consultation with a licensed mental health professional. If you are in crisis, please contact emergency services in your country immediately.
Crisis lines: US — 988 Suicide & Crisis Lifeline · UK / Ireland — Samaritans 116 123 · EU — Befrienders Worldwide
Last reviewed: May 2026.
References
- Linardon, J., Cuijpers, P., Carlbring, P., Messer, M., & Fuller-Tyszkiewicz, M. (2019). The efficacy of app-supported smartphone interventions for mental health problems: a meta-analysis of randomized controlled trials. World Psychiatry, 18(3), 325–336. doi:10.1002/wps.20673
- Goldberg, S. B., Lam, S. U., Simonsson, O., Torous, J., & Sun, S. (2022). Mobile phone-based interventions for mental health: A systematic meta-review of 14 meta-analyses of randomized controlled trials. PLOS Digital Health, 1(1), e0000002. doi:10.1371/journal.pdig.0000002
- Firth, J., Torous, J., Nicholas, J., Carney, R., Pratap, A., Rosenbaum, S., & Sarris, J. (2017). The efficacy of smartphone-based mental health interventions for depressive symptoms: a meta-analysis of randomized controlled trials. World Psychiatry, 16(3), 287–298. doi:10.1002/wps.20472
- Firth, J., Torous, J., Nicholas, J., Carney, R., Rosenbaum, S., & Sarris, J. (2017). Can smartphone mental health interventions reduce symptoms of anxiety? A meta-analysis of randomized controlled trials. Journal of Affective Disorders, 218, 15–22. doi:10.1016/j.jad.2017.04.046
- Torous, J., Bucci, S., Bell, I. H., et al. (2021). The growing field of digital psychiatry: current evidence and the future of apps, social media, chatbots, and virtual reality. World Psychiatry, 20(3), 318–335. doi:10.1002/wps.20883
- Larsen, M. E., Huckvale, K., Nicholas, J., Torous, J., Birrell, L., Li, E., & Reda, B. (2019). Using science to sell apps: Evaluation of mental health app store quality claims. npj Digital Medicine, 2, 18. doi:10.1038/s41746-019-0093-1