Promoting engagement with a mental health smartphone app: evidence from a randomized field experiment

Money can nudge students toward a mental health app, but it cannot make them stay. That is the sobering takeaway from one of the most rigorous real-world tests yet of how young people respond to incentives for digital mental health tools. In a randomized field experiment involving 1,812 university students in Sydney, researchers at the University of Technology Sydney found that paying students a small amount to download and use a well-evidenced mental health app lifted take-up by only 7 percentage points, from a baseline of 14 percent, and left engagement across all groups extremely low. The findings, published in Frontiers in Behavioral Economics, cut against the optimism that has surrounded smartphone-based mental health interventions and offer a rare, unvarnished look at what happens when such tools are offered to an entire student population rather than to a hand-picked clinical sample.

The scale of the problem the researchers set out to address is hard to overstate. Roughly one-third of university students suffer from anxiety and depression, and mental ill-health in this group is linked to reduced academic performance and higher dropout rates. Mental health apps have long been touted as a solution: they are cheap, infinitely scalable, available around the clock, and theoretically well suited to a tech-savvy generation. Hundreds of randomized trials have shown that these apps can reduce symptoms of anxiety and depression. But there is a catch that the trial literature tends to obscure. Those studies typically recruit people who are actively interested in treatment and monitor them closely to ensure they keep using the app. In real-world data, take-up and engagement are often dismal. With an estimated 10,000 to 20,000 mental health apps already on the market, the question is not whether such tools can work in principle, but whether anyone will actually use them when nobody is watching.

To find out, the team, led by economists Adeline Delavande and Nathan Kettlewell together with colleagues Beate Mueller, Rachael Murrihy, and Sonal Singh, invited all approximately 45,000 students at their university to join a study on mental wellbeing in March 2023. The 1,812 students who completed the baseline survey were broadly representative of the student body: 60 percent women, 30 percent men, 10 percent non-binary or undisclosed, with an average age of 23, roughly a quarter postgraduate, and about 30 percent international students. Crucially, the study did not restrict participation to students with clinical diagnoses. Because investment in mental health is universally important, the researchers wanted to observe behavior across the entire population, including the mentally healthy. The clinical picture they uncovered was striking nonetheless: about 36 percent of participants reported moderate to severe anxiety symptoms on the GAD-7 scale, and roughly 40 percent reported moderate to severe depression symptoms on the PHQ-9, figures consistent with international evidence.

Participants first completed a short survey, with a median response time of 6.3 minutes, measuring depressive and anxiety symptoms. They were then randomly allocated, stratified on sex, first-in-family university attendance, and first-year status, to one of four treatment arms. The control group simply completed the survey. The baseline group received a download code for MoodMission, an Australian-developed app grounded in cognitive behavioral therapy, providing 5 months of free access that covered the AUD$7.99 subscription fee. The financial incentives arm received the same code plus an AUD$5 gift voucher for downloading, and, for every mission completed, a ticket into a semester-end prize draw with AUD$500, AUD$200, and AUD$100 vouchers at stake, capped at two tickets per week. The social influence arm was invited to join an exclusive Instagram group moderated by three student ambassadors who posted short weekly videos about their experience with the app. The study was pre-registered on the Open Science Framework, and the analysis followed the pre-analysis plan.

MoodMission itself is a deceptively simple piece of behavioral technology. Users select how they are feeling, choosing among feeling low or depressed, anxious or worried, or neither, and answer a brief set of follow-up questions. The app then offers a choice of five missions drawn from a databank of evidence-based CBT strategies: behavioral activation, relaxation, mindfulness, physical exercise, and cognitive reframing. A mission might be a guided meditation, a set of push-ups, or calling a friend. After completing one, users debrief on how their mood changed and rate the activity’s usefulness, and those responses feed the app’s algorithm to tailor future suggestions. The app was chosen because it has a solid evidence base, including positive usability assessments and demonstrated improvements in wellbeing, and because it is designed to help across the spectrum from everyday mood disturbances to clinical presentations.

The results were unequivocal in one direction and humbling in another. Using MoodMission’s own user data, the financial incentive treatment increased the probability of downloading by 7 percentage points, significant at the 1 percent level, against a baseline take-up of 14 percent. When self-reported downloads were combined with the app data, the effect shrank to 5 percentage points and fell to marginal significance. Payments also increased the number of missions completed, but the absolute numbers tell the real story. Over the entire 12-week trial, the mean total number of missions completed in the baseline group was 0.17. Only 7.8 percent of baseline students completed at least one mission overall, compared with 57 percent among those who had downloaded the app, meaning that even among downloaders, nearly half never completed a single activity. In the first 2 weeks after downloading, baseline users undertook 0.07 missions per week on average, and the incentive treatment added only 0.05 missions to that figure. Conditional on downloading, the median number of missions completed over the whole trial was one.

The social influencer treatment, meanwhile, had no effect on any outcome measure. The treatment effect estimate was essentially zero, and in some specifications slightly negative. The likely reason is compliance: only 23 of the 468 students invited actually joined the exclusive Instagram group, despite several reminders and opportunities. The ambassadors were relatable peers rather than celebrities, and the opt-in nature of the approach meant the intervention never generated the exposure it needed. The researchers are careful to note that this does not mean social influence cannot work. Celebrity endorsements and messaging by influential community members have improved health literacy and vaccine uptake in other contexts, and the failure here may stem entirely from the challenge of getting students to join the group in the first place. It is a reminder that social engineering interventions face their own take-up problems, a kind of recursive version of the very issue they are meant to solve.

To understand the reluctance, the researchers turned to the students’ own words. Of 213 students who did not download the app and were asked why, 198 provided open-text answers. The dominant theme, cited by 25 percent, was perceived lack of need, with statements like my mental health is good and I feel like I’m in a good spot at the moment. Another 21 percent cited lack of time or perceived effort, and 19 percent expressed skepticism about mental health apps or a preference for other forms of help, with comments such as I don’t really believe in mental health apps and prefer going therapist. General disinterest or aversion to apps accounted for 15 percent, lack of awareness or forgetfulness for 14 percent, and technical constraints like insufficient phone storage for 7 percent. Among the 120 downloaders who explained their motivation, curiosity and general interest led at 41 percent, followed by mental health improvement at 36 percent and the free access or financial incentive at 21 percent. The themes were identified with the assistance of generative AI and then independently cross-checked and categorized by two study authors.

The theoretical implications are subtle. Motivation crowding theory warns that external interventions, particularly those perceived as controlling, can undermine intrinsic motivation. Self-determination theory, by contrast, distinguishes controlled from autonomous forms of extrinsic motivation, arguing that well-internalized extrinsic motivation can function much like intrinsic motivation when it supports basic psychological needs for autonomy, competence, and relatedness. This study was not designed to adjudicate between these frameworks, but its findings are suggestive. Extrinsic rewards did quantitatively motivate engagement, yet their effects were small, and the people who actually engaged tended to cite intrinsic motivators like wanting to improve their mood or monitor their resilience. People appear willing to download and use mental health apps for both intrinsic and extrinsic reasons, but sustained engagement remains a challenge that simple financial reward structures do not easily address, especially given that ongoing payments required for sustained use could become prohibitively costly at scale.

There is one genuinely promising thread in the data. Students experiencing mental distress were more likely to download the app, and, according to the qualitative evidence, did so specifically to improve their mental health. Women were 9 percentage points more likely to download, as were students with above-median anxiety scores by 11 points and above-median depression scores by 12 points, though the researchers found no strong evidence of heterogeneous treatment effects across the subgroups formally tested. That finding matters because previous research had suggested people in worse mental health respond less to financial incentives in small-stakes tasks. For universities and policymakers, the practical message is that if apps are to play a meaningful role in student mental wellbeing, more needs to be done to convince students of the value of the training, to address their perceived lack of time, to reassure them about the effort involved, and to remind those who simply forgot. Information about the returns to good mental health has been shown to increase willingness to pay for such apps, and reminders by text, email, or even phone could recover the forgetful. The small subset of students who actively used MoodMission may well have benefited from it. The larger lesson, however, is that in the uncontrolled reality of an entire student population, a good app and a small payment are not enough.

Subject of Research: Promoting engagement with a mental health smartphone app: evidence from a randomized field experiment

Article Title: Promoting engagement with a mental health smartphone app: evidence from a randomized field experiment

Article References: Delavande, A., Kettlewell, N., Mueller, B., Murrihy, R., & Singh, S. (2026). Promoting engagement with a mental health smartphone app: evidence from a randomized field experiment. Frontiers in Behavioral Economics, 5, Article 1700973. https://doi.org/10.3389/frbhe.2026.1700973

Image Credits: AI Generated

DOI: 10.3389/frbhe.2026.1700973

Keywords: Promoting, engagement, mental, health, smartphone, evidence, randomized, field, experiment, scientific research