This SRMA followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines. The protocol for SRMA was registered with the International Prospective Register of Systematic Reviews (PROSPERO) with registration ID CRD420251052442.

The primary objective was to evaluate the effectiveness of various interventions in managing problematic use of digital technology across domains such as internet use, smartphone use, gaming, social media use, pornography watching, gambling, OTT content watching, shopping or buying, and screen time. Secondary objectives included comparing the effectiveness of different intervention modalities; examining differential effects across age groups and conditions; and identifying methodological and quality gaps in the evidence base.

Electronic searches were conducted in PubMed, Scopus, and Embase for studies published up to April 2025. The search was conducted in May 2025.

Problematic use of digital technology encompassed the use of digital devices (online and offline) and the internet. It included various contexts including the use of the internet, smartphones, gaming, social media, gambling, shopping or buying, OTT content watching, pornography watching, and excessive screen time. Problematic screen time was used as a time-based construct only when it was measured independently and not embedded within a specific behavioral diagnosis. A comprehensive set of search terms combined controlled vocabulary (eg, MeSH [Medical Subject Headings]) and free-text keywords covering both the condition and intervention components. The search string included terms such as: (“excessive internet*” OR “problematic internet*” OR “internet addiction” OR “internet gaming disorder” OR “gaming disorder” OR “internet gaming disorder” OR “social media addiction” OR “pornography addiction” OR “smartphone addiction” OR “online gambling” OR “problematic online gambling*” OR “excessive screen time”) AND (“intervention*” OR “therapy” OR “treatment” OR “training” OR “program*” OR “effectiveness”).

The detailed search strings are available in Table S1 in Multimedia Appendix 1. The search targeted the title, abstract, and keyword fields. Manual searches of reference lists of relevant reviews and included studies were also conducted to identify additional articles.

All search results were imported into Rayyan software for deduplication and screening. Two reviewers (RKB and RS) independently screened titles and abstracts against eligibility criteria. Full-text screening was subsequently performed by the same reviewers. Discrepancies were resolved by discussion. If needed, a senior author (YPSB) resolved the conflicts. Reasons for exclusion were documented at each stage, and the process was summarized in a PRISMA 2020 flow diagram. Additionally, an interrater reliability analysis was conducted during the initial screening process to assess the validity of the studies included in the full-text screening and data extraction.

Data were extracted independently by two reviewers (RKB and RS) using a standardized form developed in Microsoft Excel. Extracted variables included study identification (author, year, and country); design and sample characteristics (n, age, and gender distribution); intervention details (type, duration, format, and theoretical basis); comparator type (active, waitlist, or no-treatment control); outcome measures; result, effect sizes, or pre-post mean differences. Disagreements were resolved by consensus.

The Cochrane Risk of Bias 2.0 (RoB 2) tool was used for RCTs, and the Risk Of Bias In Non-Randomized Studies - of Interventions (ROBINS-I) tool for nonrandomized and quasi-experimental studies. Each domain (randomization, allocation concealment, blinding, outcome completeness, and selective reporting) was rated as low, moderate, high, or critical risk.

Given anticipated heterogeneity in intervention modalities and outcome measures, both qualitative synthesis and meta-analysis were conducted. The meta-analysis was performed using Jamovi (version 2.5) [14]. Descriptive synthesis summarized intervention characteristics and results by intervention type, target behavior, and population. For studies reporting comparable quantitative data, standardized mean differences (SMDs) with 95% CIs were calculated. Random-effects models (DerSimonian–Laird) were used due to expected heterogeneity. Analyses were run for treatment-versus-control comparisons and for pre-post change within treatment arms. Statistical heterogeneity was assessed using the I² statistic and Cochran Q test. Subgroup analyses were pre-specified for type of intervention (psychological, digital, pharmacological, combined, etc); and type of problematic use of technology. Publication bias was evaluated using funnel plots and Egger regression test. In case an intervention was offered using a digital platform, it was categorized as a digital intervention irrespective of the nature of the intervention offered.

The study is based on the previously published data and did not require ethical approval.

The initial search retrieved a total of 6102 records, using the databases of PubMed (n=4305), Scopus (n=978), and Embase (n=819). After removal of duplicates, 1206 studies remained for title and abstract screening. Overall, 125 studies [15-139] were included in the systematic review (Cohen κ value of 0.98 for interrate agreement), including RCTs (n=73) [15-17,21-24,26,31,35,37,38,40-43,47,48,50,51,54-56,58,62,63,66,67,70-72,74,75,78-81,83-85,87,91,92,95-100,106-108,111-114,116-118,120,122,124-127,131,134,135,137,139], non-RCTs (n=32) [18,33,34,36,44,45,49,52,53,59-61,64,65,76,77,88,89,93,94,101,102,105,115,119,121,123,125,130,132,133,136], pre-post single-group interventions (n=14) [28,29,32,39,57,61,69,78,86,93,101,116,138], and pilot or feasibility studies (n=6) [30,41,68,86,103,129]. Collectively, these studies included more than 8000 participants. Study publication years ranged from early 2000s to 2025(Multimedia Appendix 2) (Figure 1).

The studies were conducted across various countries including China (n=39) [16,19,25,26,30,38,41-43,45,47,48,54,62-64,66-68,70,71,73,74,83,95,98,100,103,107,114,116-118,122,123,127,129,131,139], South Korea (n=15) [46,50,58,59,65,69,85,111,120,121,125,126,133,135,140], Iran (n=11) [18,20,33,40,87,89,96,97,101,132,136], Germany (n=9) [34,36,37,91,92,99,106,137,138], United States (n=9) [39,75-77,84,93,110,113,131], Turkey (n=7) [17,51,72,88,105,108,119], India (n=6) [27,29,35,49,61,94], Australia (n=3) [52,56,81], Switzerland (n=3) [21,55,134], Indonesia (n=2) [28,90], Israel (n=2) [15,128], Sweden (n=2) [78,104], and United Kingdom (n=2) [31,79]. One study each was published from Bosnia and Herzegovina, Brazil, Egypt, Finland, France, Greece, Indonesia, Malaysia, Japan, Latvia, Nigeria, Norway, the Philippines, Spain, and Thailand.

The interventions were implemented in diverse settings including schools (n=31) [15-17,20,26-29,33,48-51,65-70,82,85,88,89,96,105,109,115,123,126,132,133], colleges or universities (n=36) [18,19,22-24,30,34,38,40,47,53,58,66,68,70,72,73,75,76,83,90,92,96,100,102,112,116,117,119,125,127,130,131,136], hospitals or clinical settings (n=31) [34-37,43,44,46,54-56,60-64,77,78,81,92,93,102,110,111,113,114,118,120-122,124,125,135], digital platforms (n=17) [56,81,84,91,94,95,97,99-101,103,106-108,113,137,138], laboratory or field (n=3) [75,122,134], and community settings (n=5) [70,71,74,118,123]. Participants included adolescents, university students, and adults aged between 12 and 50 years, with sample sizes ranging from 6 to above 800 participants and intervention durations of 2 to 24 weeks.

Across all intervention categories, various behaviors in the context of problematic use of technology were targeted including internet addiction (IA) or problematic internet use (PIU) or internet use disorders (IUD; n=52) [15-43,75-77,91-102,110,114,115,123,128,131,136-138], internet gaming disorder or problematic gaming or problematic online gaming or gaming disorder (n=35) [44-64,85-87,103-105,111,112,120-122,125,130,131], smartphone addiction or problematic smartphone use or mobile phone dependence (n=23) [65-74,83,88-90,106,107,116-118,126,127], problematic gambling or gambling disorder (n=5) [78-81,113], problematic screen time or excessive screen time (n=4) [84,109,119,129], social media addiction or problematic social media use (n=3) [82,83,108], and problematic pornography watching (n=3) [75-77].

Six studies [85-90] tested educational programs that aimed to raise awareness, teach self-regulation, and promote healthier digital habits. Interventions were school- or university-based and ranged from a single-day workshop to multiweek counseling with follow-up messages. Outcome measures included the Smartphone Addiction Scale for Adolescents (SASA), IAT, Young Internet Addiction Scale (YIAS), and measures of readiness to change and self-efficacy.

Twenty-two studies [91-109,136-138] evaluated digital or technology-delivered interventions, including web-based CBT, online ACT, virtual reality (VR) modules, telehealth psychotherapy, online meditation, photography-based reflection tasks, digital education, and app-based mindfulness or self-monitoring tools. Program duration ranged from 10 days to 12 weeks. Most interventions used asynchronous modules, telehealth guidance, or app-based practice tasks.

Four studies evaluated pharmacological agents for internet, gaming, and gambling-related problems. These trials tested antidepressants (bupropion and escitalopram), antipsychotics (olanzapine), stimulants (methylphenidate), and noradrenergic agents (atomoxetine). Treatment duration ranged from 7 to 12 weeks. Outcomes were assessed using the YIAS, Clinical Global Impression – Improvement Scale (CGI-I), Bergen Internet Addiction Scale (BGCS), PGSI-related scales, and psychiatric symptom measures.

Seven studies evaluated exercise-based interventions aimed at reducing problematic technology use. Programs included aerobic exercise, tai chi, Qigong, structured physical activity, treadmill training, and recreational sports. Intervention lengths ranged from 8 to 15 weeks, with sessions delivered 2 to 3 times per week.

Five studies evaluated multicomponent interventions integrating psychological, behavioral, and physical elements. These programs combined modalities such as CBT, narrative counseling, psychoeducation, mindfulness, and structured exercise. Interventions ranged from 5 to 10 sessions delivered over 6‐8 weeks and targeted multiple mechanisms, including impulsivity, emotional regulation, craving, attentional control, and functional impairment.

Two studies assessed parent-focused interventions aimed at reducing problematic internet use and excessive screen time in children. These programs targeted parental behavior, family communication, and home-based digital regulation practices. Interventions ranged from structured parent-training sessions to multicomponent community programs delivered in a cluster-randomized design.

Treatment–control differences were significant (estimate= −1.01, SE 0.42, P=.02). Heterogeneity was high (I²=91.57%). Pre-post change did not meet significance (estimate=2.51, SE 1.41, P=.07; 95% CI −0.24 to 5.27), with marked heterogeneity (I²=96.36%; Figure 4).

Evidence within these groups was limited and did not support pooled estimates.

The overall methodological quality of the included studies varied considerably across intervention categories. Based on Cochrane RoB 2 (for RCTs) criteria, 44 (61.9%) studies [15,21,26,30,35,38,40-43,48,58,63,66,72,74,75,81,84,85,87,92,97,98,104,106-108,112,114,116-118,120,122,124,126,127,131,134,135,137,139] were rated as having some concern, 19 (26.7%) studies [23,24,37,38,47,54-56,62,71,75,79,91,96,100,111,113,125] as low risk, and 8 (11.2 %) [17,22,31,42,51,80,95,99] as high risk of bias. Most RCTs demonstrated generally low risk of bias across the majority of domains (Figure 7). Bias arising from the randomization process, bias due to missing outcome data, and bias in measurement of the outcome were predominantly rated as low risk, indicating that random allocation, retention, and outcome assessment procedures were typically well controlled. However, a notable proportion of studies showed “some concerns” in the domains related to deviations from intended interventions and selection of the reported result, suggesting incomplete reporting of adherence, protocol deviations, or prespecified analysis plans. Only a small number of studies were judged to have a high overall risk of bias, mostly due to issues with outcome reporting or insufficient methodological transparency (Figure 8). Overall, the evidence base from randomized trials can be considered methodologically robust, although interpretation should account for selective reporting and incomplete intervention fidelity in some studies (Figure 9).

The 2 RCTs [82,109] showed predominantly low risk of bias across most assessed domains. Both trials demonstrated low risk in the randomization process, missing outcome data, and outcome measurement, indicating appropriate allocation procedures and reliable outcome reporting. However, Manwong et al [82] exhibited high risk in deviations from the intended intervention, likely due to insufficient adherence reporting or uncontrolled protocol deviations. Additionally, both studies showed “some concerns” in the timing of participant recruitment and in the selection of reported results, suggesting potential issues with prespecification or full disclosure of outcomes. Overall, both trials were rated as having some concerns, reflecting generally sound methodology but with minor limitations that should be considered in interpretation of results.

Based on Cochrane ROBINS-I (for quasi-experimental and pre-post design studies) criteria, 1 (2.1%) study was rated as low risk, 20 (43.4%) as moderate risk, 20 (46%) as serious risk, and 5 (10.8%) as critical risk of bias (Figure 10). The most common sources of bias were related to selective outcome reporting and confounding. Several studies failed to report all predefined outcomes or lacked a prespecified analysis plan, resulting in moderate to serious risk in the “selection of the reported result” domain. These studies additionally demonstrated issues with confounding due to inadequate adjustment for baseline differences. Some studies also showed concerns in participant selection and deviations from intended interventions, largely due to limited methodological reporting. These limitations suggest caution when interpreting effect estimates, especially in studies with elevated risk profiles (Figure 11).

This SRMA used a cross-domain frame that treated problematic use of digital technology as a set of related conditions shaped by shared mechanisms but expressed across diverse platforms. It brings together evidence from psychological, digital, behavioral, pharmacological, neuromodulation, physical activity or exercise-based, multicomponent, parental, and VR-based interventions. It included online and offline digital behaviors and spanned age groups and settings. This broad approach allowed direct comparison across intervention types and conditions and helped clarify which approaches show consistent benefit, which remain uncertain, and where the field needs more research. The present work provides a comparative summary of the evidence and provides meta-analytic comparisons and gives the readers a broader global picture of interventions for problematic use of digital technology, while previous literature looked at individual comparisons.

Psychological interventions formed the largest group and showed the most consistent benefit. CBT, ACT, mindfulness, DBT, and skills-based programs reduced symptoms across internet addiction, gaming disorder, smartphone addiction, excessive screen time, pathological gambling, and pornography addiction. Digital and physical activity or exercise-based interventions also produced clear gains, although effects varied across studies. Combined programs that linked behavioral and physical or psychoeducational components showed broad improvements despite their smaller number. Evidence for pharmacological, neuromodulation, VR-based, educational, and parent-focused interventions was limited but showed promise in select contexts.

Meta-analysis showed strong pooled effects for psychological, digital, and exercise-based programs for internet addiction. Psychological and exercise-based interventions were effective for smartphone addiction. Pooled estimates for gaming disorder did not reach significance. High heterogeneity was present in most models.

The current findings are in keeping with the earlier reviews showing that psychological interventions offer benefit in the context of specific addictive behaviors [141]. Previous research noted that CBT reduces maladaptive thoughts and helps users regulate cues and craving [142,143]. Our results support this view. Studies in this review showed that CBT improves control of use and reduces distress across multiple conditions. Mindfulness-based care has shown similar results in earlier work. Li et al [131] found that mindfulness reduces craving and enhances attention control in internet addiction, a pattern mirrored in our findings.

Digital interventions offer wide reach and have gained attention in recent years. Andrews et al [144] reported that web-based CBT and guided self-help improve depressive symptoms when adherence is high. Our findings show a similar trend. Digital programs that used reminders, guidance, or steady practice produced stronger gains than unguided formats. These results suggest that digital care may support scale and access if programs ensure dose, guidance, and engagement. This is of particular relevance in the low- and middle-income countries where uptake of digital interventions for mental (including addictive disorders) is limited [145].

Exercise-based interventions are an emerging area. Weinstein and Lejoyeux [146] proposed that exercise may offset stress and improve self-regulation in those with high digital use. Our review supports this idea. Aerobic activity, yoga, and tai chi reduced craving and improved mood and attention in groups with problematic use of digital technology in the form of the internet and smartphone. These effects were strongest in structured, repeated sessions.

Psychological interventions were found to have a larger effect compared to exercise-based interventions. The findings suggest that strategies focusing on behavioral modifications and physical regulations would benefit from inclusion interventions addressing cognition to mitigate problematic use of digital technology. This aligns with the concept of behavioral addiction, suggesting that problematic use of digital technology involves impairment in impulse control, cognitive distortions, and emotional dysregulation [147].

Pharmacological interventions had mixed support in prior work. The current findings support this view. Antidepressants reduced gaming symptoms in some trials, but many gains were in the context of the mood change rather than change in pattern of use. Stimulants, antipsychotics, and noradrenergic agents had limited effect. Neuromodulation studies showed short-term change in craving and gaming patterns.

Educational and parent-focused interventions showed modest benefit. This is in keeping with the prior research showing that knowledge and education, while important, may not be sufficient on its own to lead to sustained behavior change [148]. In our review, behavior improved only when programs added active tasks, self-monitoring, or steady home involvement. VR-based care has seen early use in exposure therapy for gambling and gaming. VR can reduce some symptoms in the short run, but durability is unclear.

Multicomponent approaches appear promising. Previous reviews argued that problematic digital use is complex and may require programs that target emotion, behavior, and context together [149].

The combination of high heterogeneity, small sample sizes in several intervention groups, and significant Egger tests indicated that results should be interpreted with caution. Selective reporting and methodological variability may have influenced the magnitude of observed effects.

Evidence for problematic use of technology in the context of social media use, pornography watching, gambling, shopping or buying, OTT content watching, and problematic screen time remains limited. The limited research focus on therapeutic interventions for addictive behaviors has been reported previously [150]. Moreover, some digital technology use-related behaviors such as OTT content watching have generated limited research interest [151].

These results suggest that interventions that address linked cognitive, emotional, and behavioral factors in a structured way, with steady practice and clear guidance, produce the most stable gains. These findings also support the use of digital and exercise-based methods to widen reach.

This review has several clear strengths. This SRMA adds value beyond earlier work by covering all major forms of problematic use of digital technology in one analysis. Prior reviews focused on single behaviors such as internet addiction [152], gaming disorder [149], or pornography watching [153]. Others limited their scope to one age group, usually adolescents or university students, or to one type of intervention such as CBT [149]. Such narrow scopes helped define early patterns but did not capture how interventions perform across conditions. The concept of generalized and specific internet and digital device use related behaviors continues to be debated in literature [154]. Similarly, recommendations have been made to distinguish essential and nonessential internet use [155]. The current SRMA uses a broad frame that presents integrated synthesis of evidence on problematic use of digital technology. This allowed direct comparison across various behaviors in the context of problematic use of digital. The review covers all major types of interventions, including psychological, educational, digital, pharmacological, exercise or physical activity, VR, neuromodulation, and combined interventions. The review includes both online and offline forms of digital technology use and covers work with adolescents, young adults, and older adults. This wide scope gives a full view of current treatment options. The studies span many regions, which improves the reach and real-world value of the findings. The analysis used standard checks for bias, which adds to the strength of the pooled results. It also points out areas with few studies, such as OTT content watching, shopping or buying, and pornography watching. This synthesis of evidence can help strengthen the earlier recommendations on safe and healthy use of technology [156].

Several limitations affect interpretation of this SRMA. Many studies had small samples and short follow-up. Most relied on self-report measures rather than logs or device data. Intervention dose and fidelity were not always clear, and adherence was rarely tracked in detail. Study quality was mixed. Risk of bias was moderate in many nonrandomized studies, with common issues in outcome reporting, confounding, and documentation of protocol adherence. High heterogeneity reflects variation in intervention format, intensity, and outcome tools. Several nonrandomized studies showed bias due to weak control of baseline traits, unclear steps in the intervention, or gaps in outcome reports. Even among randomized trials, some did not report plans for analysis or checks on how well the intervention was delivered.

It is important to note that the small number of gambling studies included in this SRMA reflects inclusion of only digital or online gambling interventions. A substantial body of literature exists for land-based gambling disorder. However, these studies were beyond the scope of this review [157].

The pooled results also showed high variation across studies. Trials differed in target groups, treatment dose, number of sessions, and the tools used to measure change. This wide spread lowers the strength of the pooled estimates. Tests for small-study effects showed signs of bias in several groups, and this may raise the size of some pooled effects. Some areas, such as social media use, pornography use, gambling, and screen time, had too few studies for pooling. There were no studies on OTT content watching. Most studies came from school or college settings in a few regions. As a result, the findings may not apply to other settings or broader groups.

The findings of this SRMA show that structured care can reduce problematic use of digital technology. Psychological interventions, such as CBT, ACT, and mindful practice, led to the most steady improvements across conditions. Digital interventions also reduced symptoms when they offered clear steps, brief guidance, or steady practice. This shows that remotely delivered care can reach more people when support is built into the program. Exercise-based work lowered symptoms in several groups and may serve as a useful add-on for people who respond to routine, movement, or changes in mood and focus. Blended models that join psychological interventions with movement or mindful tasks may help by acting on more than one driver of problem use.

The findings also point to gaps in the research that need attention. Studies used many tools, short follow-up, and mixed forms of interventions. A shared set of measures and longer follow-up after treatment would make evidence clearer. Strong evidence is still missing for social media use, pornography watching, gambling, OTT content watching, and screen time. Few studies tested adults in clinic settings, so more work is needed beyond schools and colleges. The findings have specific implications in the context of low- and middle-income countries. Despite well-documented problematic use of digital technology from these countries, the studies exploring the therapeutic interventions remain limited [158-161].

Future studies should include larger samples, longer follow-up, and standard outcome sets. Clear reporting on intervention dose, adherence, and prespecified plans is needed. Studies should test stepped-care models and evaluate combinations of psychological, digital, exercise, and family-based components. Evidence for adults, clinical populations, and users with high distress is limited and requires expansion.

Across the available evidence, psychological interventions offer the most reliable benefit. Digital and exercise-based interventions also show value and can broaden reach. Combined models may offer added strength by addressing cognitive, emotional, and behavioral drivers together. Pharmacological, neuromodulation, VR-based, and parental interventions show early promise but need further study. Evidence for problematic use of technology in the context of social media use, pornography watching, gambling, shopping or buying, OTT content watching, and problematic screen time remains limited.

Many forms of care can reduce problematic use of digital technology, but sustained change depends on structured tasks, guided support, and continued practice. Programs that follow these principles can help individuals gain control, reduce harm, and restore healthier patterns of technology use.