In 2022, nearly 60% of the world’s population spent an average of 2 hours and 27 minutes per day on social media platforms [1]. While these platforms offer various positive benefits, like providing a sense of closeness or inspiration [2], they also possess [1] features that can foster habitual engagement and contribute to cycles of instant gratification, thereby potentially impeding the pursuit of long-term goals [3].

This duality has spurred an expanding body of research examining the complex relationship between social media use and well-being, highlighting the importance of individual, contextual, and content-related factors in shaping its impact [4-9]. Although findings on the relationship between total time spent on social media and well-being remain inconsistent [7,10], problematic social media use, characterized by excessive, compulsive, or poorly regulated engagement that disrupts daily functioning, has been consistently associated with reduced well-being [7,11,12]. In particular, problematic social media use has been associated with adverse health outcomes, including reduced life satisfaction, increased symptoms of depression, loneliness, and stress [11], as well as lower levels of physical activity [9] and decreased productivity [13]. A meta-analysis by Cheng et al [14] further illustrates the global relevance of this issue, estimating that the prevalence of problematic social media use among adults and adolescents reaches up to 31% in the most affected countries.

A healthy balance between connection and disconnection is a key aspect of “digital well-being,” which has been defined as an “optimal balance between the benefits and drawbacks obtained from mobile connectivity” that is achieved “when experiencing maximal controlled pleasure and functional support, together with minimal loss of control and functional impairment” [15]. Problematic social media use, on the other hand, is characterized by a compulsive pattern of social media use that interferes with daily functioning [16]. It displays elements of addictive behaviors, including cognitive salience, mood modification, tolerance, withdrawal, conflict, and relapse [17,18].

The Integrative Model of Mobile Media Use and Need Experiences (IM³UNE model) highlights self-control—the ability to override or inhibit immediate desires in favor of goal-directed behavior [19]—as an important factor in the relationship between mobile connectivity and digital well-being. High levels of self-control protect users from excessive or habitual social media use [20], while deficits in self-control are closely linked to problematic social media use [21].

To better understand self-control in both digital and nondigital environments, research has drawn on the dual-process model. This framework posits 2 modes of information processing: an automatic, impulsive, and unconscious process (System 1), and a reflective, deliberate, and conscious process (System 2) [22]. In the context of problematic social media use, System 1 reflects the habitual, cue-driven use of social media, which is frequently associated with adverse well-being outcomes [23,24].

Specific design features of social media platforms, often referred to as “dark patterns,” are deliberately constructed to exploit System 1 processing and maximize user engagement [25]. For instance, infinite scrolling mechanisms create an illusion of endless content and contribute to compulsive engagement by removing natural stopping cues [26]. Once immersed in these apps, disengaging becomes increasingly difficult, as System 2, despite recognizing the negative consequences of such use, often fails to override automatic behavior. This interplay between automatic engagement and diminished self-regulation underscores the need for interventions that facilitate reflective, intentional technology use. As a result, digital self-control tools have emerged to help individuals regulate their social media use and regain control over their behavior [27].

Digital self-control tools aim to prompt System 2–driven behavior to support users in regaining self-control over their smartphone use. These tools, often informed by interventionist theories like nudge theory [28], guide users’ choices and behaviors in the digital environment through behavioral interventions. Many digital self-control tools operate as self-nudges, enabling users to set up choice architectures that protect their goals from competing temptations [29]. Most of these interventions often use a “just-in-time” approach to intervene precisely when overuse is likely to occur [30].

Several apps have been proposed in recent years with varying techniques and degrees of enforcement [31-36]. For instance, the app MyTime offers real-time feedback on time spent across apps and prompts goal reflection, which led to reduced daily use and increased perceived self-regulation in a randomized trial with college students [31]. MindPhone, a smartphone app that interrupts habitual unlocking with intention prompts, was found to increase awareness and reduce habitual use over 14 days, especially among users with initially higher problematic smartphone use scores [36]. One Sec delays app access with a short breathing pause and prompt. A within-subject field study showed that this design significantly reduced Instagram use and increased perceived control without notable user frustration [35].

Despite promising developments, existing digital self-control tools continue to face critical limitations regarding long-term effectiveness and sustained user engagement [37-39]. Many tools rely on overly restrictive mechanisms, such as app blocking, which can undermine users’ sense of autonomy and trigger psychological reactance—ultimately reducing continued use. Others adopt low-friction approaches, such as passive notifications, which are often ignored at the moment of temptation and fail to interrupt habitual use [37,40-42].

A further limitation lies in the lack of customization: most tools do not allow users to align the intervention with their personal goals, definitions of distraction, or preferred self-regulation strategies. A systematic review of 334 digital self-control tools found that few support the nuanced, user-defined boundaries essential for meaningful and sustainable behavior change [38].

Moreover, evidence on the effectiveness of many digital self-control tools remains limited, particularly with regard to addressing the underlying psychological mechanisms of problematic social media use, such as impulsivity, habitual behavior, and deficient self-control [30,39]. Although a wide variety of tools exists, few interventions integrate empirically grounded behavior change strategies with autonomy-supportive design principles. This presents a critical gap in the literature: the need for digital self-control tools that combine customization, user agency, and theory-based mechanisms of change. Studies indicate that users are more likely to engage with interventions that allow them to set individualized goals, determine when and how reminders are delivered, and track their progress toward greater self-regulation [43,44].

This study addresses this gap by evaluating the Wellspent app, a customizable, autonomy-supportive intervention designed to promote self-regulated social media use. By combining evidence-based behavior change techniques (BCTs) [45] with flexible, user-driven customization, Wellspent aims to effectively support users in regaining control over their usage patterns while preserving users’ sense of autonomy. Through a preregistered randomized controlled trial, this study offers one of the first rigorous evaluations of a digital self-control tool specifically designed to target problematic social media use in a personalized and theory-informed manner.

The increasing need for effective tools to support healthier and more intentional social media use has led to the development of the Wellspent app, a theory-informed digital self-control tool. The app represents a novel approach to digital self-regulation by combining evidence-based BCTs [45], such as goal setting, behavioral feedback, self-monitoring, and intention-based nudges, with a high degree of user autonomy and customization. Unlike many existing tools that rely on rigid blocking mechanisms or generic notifications [38], the Wellspent app allows users to define their own usage goals, select problematic apps, and personalize reminder timing, tone, and content.

A core differentiator of the Wellspent app lies in its autonomy-supportive design. Rather than enforcing restrictions, the app preserves users’ freedom to choose whether to quit or continue using an app after each prompt. This design is grounded in the dual-process model of behavior regulation [22], aiming to interrupt automatic, habitual engagement (System 1) and foster more reflective, goal-directed decision-making (System 2). Prior research suggests that such personalized and autonomy-supportive interventions may enhance message resonance, reduce psychological reactance, and improve long-term adherence [46]. With its emphasis on autonomy, customization, and theory-driven behavior change, this study addresses a critical gap in the digital well-being literature and contributes to the development of more effective, user-centered self-regulation tools.

The significance of this study lies in its methodological rigor. To date, few interventions in the field of digital self-regulation have undergone a rigorous empirical evaluation using gold-standard methods [47,48]. This study presents a preregistered randomized controlled trial with 70 participants, comparing the effects of the Wellspent app with a no-treatment control group. The design ensures high internal validity and allows for causal inferences regarding the app’s impact on behavior.

An earlier publication from the overarching study project on intervention effects examining separate outcomes found no significant effects on overall well-being, affect, or stress but observed reduced stress among users with initially high levels of problematic smartphone use in the intervention group [47]. The primary objective of this study is to investigate whether participants in the intervention group receiving the Wellspent app, compared to a control group, demonstrate reductions in problematic social media use, problematic smartphone use, and time spent on the most problematic social media app, as well as an increase in self-efficacy for self-regulating social media app use.

This study presents findings from a randomized controlled trial designed to evaluate the effectiveness of the Wellspent app, a digital self-control tool aimed at promoting self-regulated social media use on smartphones. The trial was preregistered at the German Clinical Trial Register (DRKS00031767) on May 8, 2023. The first participant was enrolled on May 22, 2023, and a sample size of 108 participants was targeted. The study was conducted between May and June 2023.

Participants were recruited via private social media channels (Facebook, LinkedIn, WhatsApp, and Instagram) and email lists for study opportunities at the Freie Universität Berlin, Germany. Prospective participants enrolled through an online registration page, where they received information regarding the study’s purpose, data protection guidelines, and the voluntary nature of participation. Informed consent was obtained from all participants prior to their inclusion in the study.

Inclusion criteria required participants to be at least 18 years of age, proficient in English, own an iPhone with iOS 16 (as the Wellspent app was exclusively available for this operating system), and regularly use social media. Social media was broadly defined as platforms enabling users to engage with content, share updates, or interact with others online (eg, Instagram, Facebook, and TikTok).

The final study sample consisted of 70 participants (47/70, 67% female; mean age 26.2, SD 5.6 years), with 35 participants allocated to each group. The remaining 70 participants were randomly assigned (1:1) by the author team to either the intervention group (n=35) or the control group (n=35) using a simple randomization procedure via Research Randomizer [49]. No blocking or stratification was used. Formal blinding was not implemented, as participants were aware of their group assignment.

Participants assigned to the intervention group received the app’s intervention features during week 2. The control group received no intervention and continued using their phones as usual. Use of the intervention app in week 3 was voluntary for the intervention group. Both groups completed follow-up assessments at the end of week 2 (T2) and week 3 (T3). At each time point, participants were prompted via email to complete the respective questionnaires, with follow-up reminders sent 24 hours later. Demographic data (including age, sex, level of education, and occupation) were collected at baseline (T1; Figure 1).

This study was approved by the Ethics Committee of the Department of Education and Psychology at Freie Universität Berlin, Germany (proposal number 009.2023; approval date May 5, 2023). Prior to enrollment, all participants provided informed consent, having received detailed information about the study’s purpose, data processing procedures, and their right to withdraw at any time. To ensure privacy and confidentiality, all data were pseudonymized at the point of data collection, and no personally identifiable information was retained. All collected data were stored on secure servers in accordance with the General Data Protection Regulations regulations. Participants were compensated with academic course credits (when they were students) and received a 1-year premium account for the Wellspent app, provided by the app’s developer; no other financial incentives were offered.

The Wellspent app is a theory-informed, customizable mobile intervention developed to support self-regulated social media use. It targets user-identified problematic apps and delivers personalized behavioral nudges. The intervention draws on BCTs from the BCT Taxonomy v1 [45], addressing psychological mechanisms associated with problematic social media use [50]. Specifically, the intervention incorporates the following BCTs [51]:

The intervention was designed according to self-nudging principles [29], enabling users to customize the app’s behavior to align with their habits, goals, and preferences. This approach preserves autonomy and supports reflective (System 2) rather than habitual (System 1) decision-making [22]. Customization features and exemplary reminder nudges are detailed in Multimedia Appendix 1.

All users in the intervention group completed the same standardized setup process:

Weekly self-report measures were collected between baseline (T1) and follow-up (T3) to assess the following outcomes: problematic social media app use, problematic smartphone use, self-efficacy to self-regulate social media app use, and participants’ screen time for their most problematic social media app.

User acceptance of the intervention was defined as the user refraining from dismissing a reminder (and thereby continuing their app use) within 1 minute of receiving the reminder nudge. The acceptance rate of the Wellspent app’s reminders was computed by dividing the instances of accepted interventions by the total number of occurrences.

Qualitative data were collected through an online survey conducted via Qualtrics at the end of the intervention period. Participants in the intervention group answered a series of 10 open-ended questions aimed at exploring their experiences with the Wellspent app (Table S1 in Multimedia Appendix 2). These questions included prompts such as “How would you describe the Wellspent app to a friend?” and “What does the Wellspent app help you with?” The questions were designed to gather insights into users’ perceptions of the app’s functionality, usability, and effectiveness, as well as to identify potential areas for improvement. A thematic analysis approach was used informally to identify key themes from participants’ responses [56].

The list of covariates included participants’ age, sex (women=1; male=0), education (1=university degree; 0=no university degree), positive affect, and screen time in minutes at baseline. Positive affect was measured with an adaptation of the Positive and Negative Affect Schedule with items such as “Please indicate the extent you have felt this way when thinking about your social media app use over the last 7 days” (1=“very slightly or not at all”; 6=“extremely”) [47].

Prior to hypothesis testing, data were screened and cleaned. Univariate outliers in screen time (z>3.29) were winsorized to 1 unit above the highest valid score with z<3.29, following recommendations by Tabachnick and Fidell [55]. To assess baseline comparability between the intervention and control groups, t tests and chi-square tests were conducted for key demographic and outcome variables. These tests also informed attrition analyses comparing retained and dropout participants.

Of the 120 persons screened for eligibility, 33 of them did not install the Wellspent app, and 17 persons did not complete the baseline questionnaire. The randomized sample consisted of 70 individuals (47 women and 23 men) with a mean age of 26.19 (SD 5.63; range 18‐45) years. Figure 1 shows a CONSORT (Consolidated Standards of Reporting Trials) flow diagram outlining the study’s design and participant allocation across the randomized controlled trial with 3 measurement points. Most participants were students or young professionals. Participants’ most chosen problematic social media app was Instagram (29/35, 82%), followed by TikTok (4/35, 10%) and Twitter (1/35, 4%). Detailed demographic data are presented in Table 1, and descriptive statistics for all outcome measures across time and condition are provided in Table 2.

Randomization confirmed no significant differences across baseline variables between groups. In the attrition analysis, we found that participants who dropped out from the study before T3 (n=18) showed higher levels of positive affect than participants retained in the longitudinal sample (n=52; nonretainers: mean 2.59, SD 0.53 vs retainers: mean 2.20, SD 0.67; P=.03). Among the total sample, a subsample of 52 participants (representing a dropout rate of 26%) completed the study by week 3, comprising 25 participants from the intervention group and 27 participants from the control group. These numbers are almost in accordance with the power analysis, which indicated a targeted longitudinal sample size of 27 persons per group. The attrition rate of the present study aligns with observations from other online-based studies [63]. Table 1 provides an overview of key demographic characteristics of the baseline sample.

Qualitative feedback was explored informally by reviewing the open-ended responses from the intervention group, collected via Qualtrics at the end of the intervention period. Responses were grouped into recurring patterns based on the researcher’s judgment, without following a formal coding or analysis protocol. Emerging themes included the following.

This randomized controlled trial investigated the effectiveness of the customizable intervention app Wellspent in supporting self-regulated social media use. The intervention produced significant reductions in daily time spent on the most problematic app (approximately 29 minutes) and reduced perceived problematic smartphone use. In contrast, no significant effects were found for problematic social media use or self-efficacy to self-regulate social media use. These findings suggest that behavioral change, in the form of reduced usage time, can occur relatively quickly through moderate, autonomy-supportive interventions, whereas cognitive and self-perception outcomes may require longer, more reflective processes.

The intervention did not significantly decrease problematic social media use. This aligns with research showing that problematic use arises from deeper, multifaceted patterns, such as emotional health [16], habit strength [64], and personality characteristics [65], making it difficult to change quickly. The intervention app may not have supplied the sustained cognitive scaffolding needed to alter participants’ perceived problematic use within 2 weeks.

The intervention group reported lower levels of problematic smartphone use compared to the control group. This suggests that behavioral reductions may have translated into a more general feeling of control over one’s device. Notably, perceived problematic smartphone use may be easier to shift in the short term than problematic social media use because it reflects global awareness (“I use my phone too much”) rather than app-specific patterns linked to social, emotional, or identity-related motives.

Self-efficacy significantly increased in both groups, but participants from the intervention group did not exceed those from the control group. Self-efficacy typically grows through repeated mastery experiences and feedback on successful behavior regulation [66]. As the intervention app focused on in-the-moment friction rather than visible progress tracking, participants may have lacked the experience of sustained success and explicit feedback loops. Similar interventions have also failed to find strong between-group intervention effects on self-efficacy [19], pointing to the need for longer interventions, richer in feedback and reflection.

The intervention produced a reduction in time spent using social media apps of approximately 29 minutes per day. This finding aligns with prior research highlighting the potential of digital self-control tools to reduce usage time when users are nudged at critical moments [35,40]. Compared to digital self-control tools with generic interventions or one-size-fits-all designs [38], Wellspent offered a tailored and user-centric alternative. The full-screen reminder appearing after a personally defined nudge interval (eg, 10 min on Instagram) created a structured moment of pause, allowing users to shift from automatic to deliberate decision-making [22]. Previous literature emphasizes that user autonomy and personalization increase intervention acceptance and long-term engagement [46]. Our results support this perspective, particularly the positive user feedback from intervention participants regarding flexibility and customization. By integrating customizable, goal-aligned nudges into real-time usage patterns, the app sets a precedent for user-centered intervention strategies in digital self-control.

Together, the findings support emerging evidence that autonomy-preserving tools can meaningfully influence digital behavior within short periods. These findings highlight three key considerations for practitioners and designers of digital self-control tools:

This study has several limitations. First, the sample size was relatively small, and self-report measures may be subject to bias, particularly regarding subjective app usage and problematic use. The target sample size of 54 participants, as determined by the power analysis, was not reached. The final analysis sample for the multilevel models on the primary outcome consisted of 46 participants, which may have limited the statistical power to detect medium effects. Second, the short intervention duration may have limited the app’s impact on self-efficacy and habitual behavior. Third, although self-efficacy to self-regulate social media use was assessed, a direct measure of self-regulation (eg, action control) was not included in the study. This omission may limit the ability to fully capture the psychological mechanisms underlying digital behavior change. Future studies should incorporate validated measures of self-regulation to provide a more comprehensive understanding of intervention effects. Fourth, the study was limited to iOS users, excluding other relevant target groups.

Future research should explore the long-term effects of customizable digital self-control tools and investigate how additional features, such as social comparison, adaptive feedback, or intention tracking, could enhance both behavioral and psychological outcomes. Mixed method designs integrating behavioral data, qualitative feedback, and longitudinal tracking could provide deeper insights into how users engage with interventions over time.

This study provides initial evidence for the effectiveness of a customizable, moderately enforced digital self-control tool in reducing social media screen time and perceived problematic smartphone use. By allowing users to define when, where, and how interventions occur, the Wellspent app demonstrates a user-centered approach to digital self-regulation. These findings support the growing consensus that flexible and autonomy-supportive digital self-control tools offer a promising path toward sustainable behavior change in digital environments.