The Promise and Peril of Mental Health Apps

Social Media

Mental health apps promise accessible care—but most lack scientific validation. Our review finds that while some apps help with depression & anxiety, the majority make unverified claims. Regulation & transparency are urgently needed to protect consumers. #MentalHealth #Science

Key Points

Sophia, a 27-year-old marketing professional in New York City, enjoys her fast-paced job, weekend dinners with friends, and quiet nights at home with her cat, Muffin. Lately, though, she has felt a constant edge of anxiety she can’t shake. A quick online screening suggests an anxiety disorder, and her doctor recommends therapy—but the next available appointment is weeks away. Scrolling through her phone one night, Sophia finds an app that promises to ease anxiety with short daily exercises. It appears under the “medical” category, which feels reassuring—but what does that label really mean? The App Store is full of similar apps, all claiming to reduce stress or lift mood. Curious, Sophia searches for scientific reviews of mental health apps. Some studies report benefits for certain apps, yet she notices that the ones shown to work aren’t the ones she can actually download. The discovery leaves her uneasy: if most apps aren’t backed by evidence, how is anyone supposed to know which to trust?

Though fictional, the story of Sophia illustrates the dilemma of millions of people who suffer from mental illness while facing long waitlists and sizable bills for care. Globally, the prevalence of mental health disorders had risen by more than 30% since 1990—even before the COVID-19 pandemic (James et al., 2018). In the United States, mental health also continues to decline amongst adults (Wright et al., 2025), while at the same time almost two-thirds of American college students report experiencing “overwhelming anxiety” each year (American College Health Association, 2019). Existing infrastructure is failing to meet this increased demand for mental health services. In the United States, less than one in three people with mental health disorders receive treatment (Olfson et al., 2016), while the inequities in the prevalence of poor mental health continue to grow (Wright et al., 2025). Can mobile apps for mental health be part of a comprehensive solution to the mental health crisis in the United States and around the globe?

In the present article, we will show that despite some promising evidence, most mental health apps marketed to consumers are unvalidated to treat mental illness. As a result, an unregulated market of mental health apps might have little effect or even exacerbate the current mental health crisis. Accordingly, we propose regulatory and other solutions to protect consumers by ensuring that they can easily discern between validated and unvalidated apps. Our focus is on smartphone apps, but we also consider web-based apps where relevant. We examine mental health apps designed to deliver treatment without input from healthcare professionals (self-guided apps) and apps that deliver app-based treatment with the guidance of a professional (provider-guided apps). Our review focuses largely on mental health app research predating the AI-chatbot revolution in 2022, though we integrate recent findings where available. These recent technological developments underscore why policy action is now more pressing than ever. While AI chatbots show promise for democratizing mental health care, they also hold the potential of exposing millions of users to untested interventions that could delay appropriate care or cause harm.

We organize the present work into three sections. First, we review the existing evidence of the efficacy and effectiveness of mobile health (mHealth) apps for mental health. Efficacy trials test whether an intervention works under ideal and controlled conditions. In contrast, effectiveness trials assess how well an intervention performs in real-world settings, where users vary in motivation, context, and adherence. In other words, efficacy tells us if an intervention can work, while effectiveness tells us if it does work when deployed outside the lab. We highlight critical gaps in knowledge, including a lack of efficacy evidence compared to established treatments like talk or drug therapy (i.e., standard-of-care treatments), a lack of evidence in underserved populations, and a lack of evidence for real-world effectiveness. Notably, most of the existing evidence comes from studies in the United States, but we do note evidence from other countries where relevant. Second, we conduct an original analysis that demonstrates that the vast majority of apps available to consumers are unvalidated. Finally, we explore the implications of our findings for public policy, proposing specific ideas to protect and inform consumers.

We propose a multi-pronged approach to ensure the efficacy and safety of mental health apps. Our solutions are primarily focused on the U.S. regulatory landscape, but the overall principles may be applicable elsewhere. First, the U.S. Food and Drug Administration (FDA) should regulate these apps as medical devices or akin to dietary supplements, necessitating stringent efficacy and safety data or explicit disclaimers. To strengthen these efforts, we suggest refining and expanding the FDA's guidelines, such as extending regulatory oversight to consumer-facing mental health apps, requiring standardized disclaimers for untested claims, and/or requiring platforms such as Apple's App Store and Google Play to adopt consistent labeling and reporting standards. Apps should be listed under the medical category of Apple's App Store, for example, only if they meet the standards for medical devices. Every app claiming to treat psychiatric conditions should be accompanied by a disclaimer that clearly indicates whether peer-reviewed research supports its efficacy claims, ensuring transparent communication to users. Second, we propose the development of accessible databases and tools that aid consumers in selecting apps based on evidence. Custom-designed AI agents that help consumers use such databases interactively to reach better evidence-based decisions could be a way to use AI in a productive way in this space. These resources promise to simplify the evaluation of scientific support for such apps and clarify their relevance to specific demographic groups.

Part 1: Review

In our review, we use evidence from meta-analyses and existing reviews of randomized controlled trials (RCTs), where participants are randomly assigned to an app treatment or a control condition and where their mental health outcomes are compared across conditions. When meta-analyses are not available, we also consider evidence from individual RCTs. Where relevant, we evaluate the strength of the evidence by specifying both the sample size and the effect size. We use Cohen's conventions of effect size, where an effect of d/g = .20 is considered small, d/g = .50 is considered medium, and d/g = .80 is considered large (Cohen, 1988). Further details on our methodology are available in our Supplementary Materials (https://doi.org/10.17605/OSF.IO/5DKV).

Part 2: Can Consumers Trust Mental Health Apps?

Thousands of wellness and mental health apps are available to consumers on the iPhone and Android app stores (Buss et al., 2024; Larsen et al., 2019), but only a small number—in the dozens— have been tested using randomized control trials, or RCTs (Linardon et al., 2019, 2025), and most of those RCT-tested apps are not directly available to consumers (Buss et al., 2024). Yet, one analysis found that while 64% of mental health apps available to consumers made claims about their effectiveness, and 44% used scientific language to make those claims, only one app provided a citation for their claims (Larsen et al., 2019).

Given the rapidly evolving landscape of mobile mental health apps, we conducted our own analysis to understand the extent to which mental health apps available to consumers are supported by evidence. Following the approach of Larsen et al. (2019), we identified consumer mental health apps by searching the Apple App Store for terms corresponding to the nine most prevalent mental disorders worldwide based on the Global Burden of Disease Study 2019 (GBD 2019 Mental Disorders Collaborators, 2022): anxiety disorders, depressive disorders, intellectual disabilities, ADHD, conduct disorders, bipolar disorders, autism spectrum disorders, schizophrenia, and eating disorders. Guided by subcategories of the DSM-5 (Diagnostic and Statistical Manual, 5^th edition, American Psychiatric Association, 2013), we included additional disorder-specific terms (e.g., agoraphobia, panic disorder). Searches were conducted on September 25, 2025, using Apple's iTunes Search API (Application Programming Interface) and limited to English-language mobile apps. The initial search yielded 2,313 unique apps. We then filtered apps to include only those explicitly referencing treatment (e.g., “therapy” or “treatment”), excluding 1,018 apps that did not meet criteria. We additionally removed 51 apps containing the word “test” in their titles (e.g., “Depression Test”) to eliminate self-assessment tools. The final analytic sample comprised 1,244 distinct mental health apps. We did not search for Android apps since the Google Play Store does not provide official code (API) for searching available Android apps.

Next, we found that about a third of the apps we identified (424 out of 1244 apps) are listed under the medical category in the Apple App Store. Though consumers may perceive such apps to provide evidence-based medical treatments, Apple classifies apps as ‘medical’ based on the claims of the marketer rather than on the presence of validated information attesting to their efficacy as medical treatments. Third, we found that 31% (391/1244) of app descriptions claim that the app is backed by scientific evidence. Specifically, an app was classified as making a quality claim if its description contained at least one of the following terms: evidence, research, or science. Yet, a search of the peer-reviewed literature revealed that only 27% (81) of the 391 apps that made empirical claims were backed by any publicly available scientific evidence. For our literature search, we used a large language model (LLM), GPT-4.1-mini (April 2025 release). See Supplementary Materials for details of our method (https://doi.org/10.17605/OSF.IO/5DKV).

Importantly, our review of the literature was quite broad, including correlational studies, experimental studies, and clinical trials (RCTs). This means that the number of apps with strong empirical support based on RCTs is likely even smaller. Indeed, an earlier review of mobile apps for depression and anxiety identified 179 apps available to consumers, but only three apps—less than 2%—had evidence from both RCTs and real-world effectiveness trials (Leong et al., 2022). Not only are commercially available apps largely unvalidated, but many apps are also not even based on established treatments. For example, CBT is a validated treatment for anxiety. One analysis of 361 free iPhone apps for anxiety and worry available to consumers showed that about three-quarters (269) of the apps were not based on CBT—the most validated treatment for anxiety (Kertz et al., 2017). If only a fraction of people download apps with validated efficacy, and most apps do not even use validated treatments, the real-world effectiveness of the mental health apps available to consumers may be both statistically and practically insignificant.

Part 3: Policy Recommendations and Other Solutions

To the extent that some people turn to unvalidated mental health apps for treatment instead of seeking validated treatments, mental health apps might not only be ineffective but also harmful: the time and effort spent on these apps represents a missed opportunity to seek treatments that could actually help. What can be done to protect consumers? Our solutions focus primarily on the approaches relevant to the regulatory landscape in the United States, but similar approaches could be applicable in other countries, as well.

One solution is for regulatory agencies, such as the U.S. Food and Drug Administration (FDA), to regulate mental health apps as they regulate other medical devices. Medical devices are defined by the FDA as devices “used to diagnose, prevent, or treat a medical disease or condition without having any chemical action on any part of the body” (Jin, 2014). This description sounds similar to what many mental health apps are claiming to be doing. If mental health apps claim to be medical devices, then they should be regulated as such. Importantly, the FDA does consider software functions provided by some mobile apps to be medical devices, but currently, the FDA focuses on mobile apps that use phone sensors, complement existing medical devices, or are used for active patient monitoring in the category of medical devices (U.S. Food & Drug Administration, 2022). Thus, the FDA's current regulatory focus excludes most mental health apps. Even classifying mental health apps as a Class I medical device—the lowest risk category, which includes devices such as bandages, hospital beds, or electronic toothbrushes—would provide some regulatory control, such as requiring app developers to keep records of complaints and report adverse effects from consumers. It could be argued, of course, that apps claiming to treat potentially life-threatening psychiatric disorders like depression are in a higher risk category than bandages and toothbrushes, which would make them Class II medical devices. In that case, regulation would include requirements for premarket authorization based on efficacy and safety data, as well as post-market surveillance studies with actual consumers (U.S. Food & Drug Administration, 2018).

Another regulatory solution is for the FDA to regulate mental health apps similarly to dietary supplements. In that case, the FDA will not need to approve each app based on clinical trials, but will require all mental health apps to include a disclaimer similar to the one found on food supplements, such as: “These claims have not been evaluated by the Food & Drug Administration.” Interestingly, the FDA already provides guidance for disclaimers that mental health apps should include, but these recommendations are currently nonbinding. In addition, the recommendations apply exclusively to prescription-only “computerized behavioral therapy devices for psychiatric disorders.” This FDA category refers to software-delivered therapy programs—not general wellness apps—that function as a medical treatment and require medical supervision, similar to how some medications require a prescription. This definition, however, excludes the majority of mental health apps available to consumers—even those making claims that they can provide treatment or those categorized as ‘medical’ in the App Store. These FDA recommendations encourage app developers to include (1) a statement about when and how to contact a healthcare provider, (2) a prominent label that the app “does not represent a substitution for a patient's medication,” (3) an explicit statement that the app has not been clinically tested or the results and methods of any clinical trials, and (4) clear identification of any functions that are not cleared by the FDA (U.S. Food & Drug Administration, 2020). A logical next step would be for the FDA to begin requiring all mental health apps marketed for the treatment or diagnosis of psychiatric disorders to include those statements.

It is important to note that the FDA does not currently have any regulatory guidance or requirements for commercial platforms that host mental health apps, such as the Apple App Store and the Google Play Store. As we saw in our analysis of apps available to consumers, this has allowed Apple to classify apps as ‘medical’ without following the same FDA guidelines for the classification and approval of medical devices. Even without explicit FDA regulation, Google and Apple could voluntarily adopt the FDA guidelines by only classifying apps under the medical category if they have met the burden of proof for medical devices. In addition, Apple and Google could require each app claiming to treat psychiatric conditions to, at the very least, be accompanied by a disclaimer, such as: “These claims have not been validated by the scientific community.” Unlike strictly regulatory solutions, such company-endorsed and company-initiated solutions have the notable advantage of influencing mental health outcomes globally, not just in the United States.

Looking beyond solutions that depend on underfunded regulatory agencies and the goodwill of for-profit corporations, consumers also need access to user-friendly central databases where they can check the evidence for the efficacy of each app. Several such tools, such as MindTools (Neary & Schueller, 2018) and MindApps (Beth Israel Deaconess Medical Center, Division of Digital Psychiatry, n.d.), are already available to consumers. MindApps, for example, is a free tool available to all, with a list of filters to tailor content. These tools provide a rating for each app based, in part, on scientific evidence. To ensure consumers are aware of these resources, funding is needed to increase consumers’ awareness and use of these databases.

Furthermore, future versions of such tools should make it even easier for consumers and providers to evaluate the strength of the existing evidence. An example of such a tool is the website examine.com, which evaluates the evidence for the efficacy of herbal and nutritional supplements. For each supplement, the website provides an evidence table that easily shows consumers how strong the evidence is for the efficacy of each supplement on each outcome, provides information on the magnitude of the effect, indicates how consistent the findings are, and supplies a list of all relevant citations. In addition, future tools should also include information on the efficacy of these apps in specific populations. Given that most studies identified in meta-analyses did not include people over 65, for example, consumers need to know that efficacy has not been established in older populations. Finally, future consumer-facing tools to evaluate the efficacy and effectiveness of apps can be enhanced by custom AI agents powered by large language models. This can help to improve the engagement of consumers with otherwise complex databases while also making it easier for providers to maintain and update these tools over the long term.

Conclusion

Overall, evidence suggests that mental health apps, including some self-guided apps, can be efficacious in reducing symptoms of some mental health conditions, primarily depression and anxiety. And recent evidence shows the promise of AI-driven chatbots in making mental health apps more engaging for consumers and efficacious in their impacts. Consumers should know, however, that the efficacy of the vast majority of apps available to them has not been established. In addition, there is very little evidence demonstrating the effectiveness of mental health apps in real-world settings. Evidence is also scant to test the efficacy of these apps in particular underserved populations, such as men, the elderly, and Black people, undermining their promise to provide better care to the underserved. Furthermore, the efficacy of digital treatments has not been tested compared to standard-of-care, such as talk or drug therapy. Thus, for now, apps and digital treatments cannot be recommended as a replacement for established in-person therapies, but they can be used to supplement treatment and offer treatment sooner when in-person therapy is not immediately available (Lecomte et al., 2020). Without further evidence and clear guidelines, however, we caution that this approach reinforces a two-tier system of mental health treatment, whereby people who can afford therapy receive standard-of-care treatment while underprivileged populations are left to rely on lesser, digital treatments.

Mental health apps could become a powerful tool in combating the global mental health crisis. To realize their true potential in democratizing mental health care, we need more evidence to determine what works and for whom, and more regulation to curtail the spread of unvalidated apps that may be doing more harm than good. There are many paths forward: regulatory agencies like the FDA can extend their oversight to mental health apps that claim to treat psychiatric conditions, requiring, at a minimum, that these apps include clear disclaimers about the evidence supporting their claims. App store platforms like Apple and Google can voluntarily adopt FDA guidelines and only classify apps as ‘medical’ if they meet appropriate standards of proof. Finally, consumer-facing databases that evaluate app efficacy can be supported with increased funding and visibility. With the right combination of rigorous research, thoughtful regulation, and accessible information, mental health apps can move from being a largely unvalidated marketplace to a legitimate component of evidence-based mental health care.

Supplemental Material