This systematic review followed reporting guidelines from PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) and SWiM (Synthesis Without Meta-Analysis), respectively [25,26]. The review was preregistered in the International Prospective Register of Systematic Reviews (PROSPERO) database under registration number CRD42023473117. The review protocol was also published after peer review [27]. No ethical approval was required as this review uses only publicly available literature.

The search strategy for the review was developed in consultation with an information specialist, and the rationale for the search strategy, including scope of dates and search terms, is reported in the additional material according to PRISMA–Search extenion (PRISMA-S) [28]. The search strategy used a combination of keywords and database-specific subject headings (eg, Medical Subject Headings terms) to capture peer-reviewed literature related to JITAIs, AYAs, digital health technologies, and health outcomes and was peer-reviewed [27]. The literature search took place across multiple electronic databases to identify relevant literature from 2013 to 2025; these databases included PubMed, Scopus, Web of Science, PsycINFO, Embase, Cumulative Index to Nursing and Allied Health Literature, Institute of Electrical and Electronics Engineers Xplore, Association for Computing Machinery Digital Library, Cochrane Central, Dimensions, World Health Organization Global Index Medicus, and Google Scholar.

The individual database searches were conducted in July 2025, with additional keyword-based searching in Google Scholar until December 2025. Reference list screening of extracted studies was used to identify further eligible studies. Further, studies included in prior systematic reviews of JITAIs were consulted [19-24], which enabled inclusion of all relevant studies published before (as the literature search strategy only started in 2013, the first explicit mention of JITAI and AYAs in the literature [29,30]). These steps helped ensure comprehensive coverage of past and emerging studies. It was not deemed necessary to contact any study authors.

All literature search records were imported into Covidence, a systematic review management platform developed by Veritas Health Innovation, with duplicates automatically removed. Each title and its accompanying abstract were screened by 2 reviewers (either KWG, CA, CI, IK, and/or RA, each independently), following a pilot screening phase to ensure consistency in eligibility decisions. Full texts were also screened by 2 reviewers each. All discrepancies during screening or assessment of full texts were resolved through discussion.

Study inclusion was primarily determined by whether studies provided sufficient detail on JITAI components to inform the review’s focus on how these interventions adapt to AYA contexts and health needs. Studies were eligible for inclusion if they were peer-reviewed, focused on AYAs aged 10-25 years, reported mHealth interventions that adapted in real time to user data, and were consistent with the defining features of JITAIs. Other mHealth studies were out of scope, including JITAI studies that involved participants beyond the AYA age range (determined by reported sample age range and mean age, where available) and did not report these findings separately [31]. We selected the age range of 10 to 25 years, given its increasing recognition as its own heterogeneous yet distinct developmental period among scholars [10,32-34]. We did not restrict by study type or design, such as control conditions. We also excluded just-in-time simulation studies that did not involve evaluating an intervention [35].

Data from each study were also extracted by 2 reviewers (either KWG, SG, VC, and/or RA, each independently) using a piloted, standardized data extraction form to ensure consistency. The data extraction form followed standardized conceptualizations of JITAI in the literature [16,27], which outline key components guiding interventions to be adapted to individual needs: intervention options or the different forms of support or methods for delivering that health support, which may be considered based on tailoring variables; tailoring variables, referring to the data collected about an individual, which inform how and when an intervention should be adapted; and decision rules or the guidelines that use these tailoring variables to determine the selection, timing, and delivery of intervention options at specific decision points or specific moments, such as particular times of the day or contextual triggers when these decisions are made [16,27]. These components together shape how JITAIs dynamically adjust to ensure interventions remain relevant across shifting adolescent contexts [16]. By structuring the form around these components, the method was aligned with the review’s focus on mechanisms of adaptability in JITAIs.

Further, distal and proximal outcomes were also extracted, as they indicate the health objectives toward which tailoring mechanisms are directed. Specifically, distal outcomes represent the overarching, long-term goals of a JITAI, while proximal outcomes are short-term health objectives that often serve as mediators that lead toward these distal targets [16]. Thus, data extracted from each study included participant characteristics, intervention details, tailoring mechanisms, study design, and outcomes reflecting the goals of tailoring mechanisms (both proximal and distal). Refer to Multimedia Appendices 1-3 for the full extraction data for each study.

No formal amendments were made to the registered or published protocol [27]. Several additional outcomes of interest, including user engagement and ethical considerations, were identified in the protocol as potentially informative for interpreting effectiveness; however, inconsistent measurement and incomplete reporting meant they could not be systematically synthesized or graded. While information on AYA engagement and ethical considerations was documented during data extraction for each study where reported, reporting was sparse and also lacked sufficient detail for structured synthesis.

Two reviewers (KWG and MA) independently assessed methodological quality, including risk of bias, using the Joanna Briggs Institute checklists for distinct study designs, with discrepancies resolved through discussion [36]. The Joanna Briggs Institute checklists for each study included critical appraisal according to temporal precedence, selection and allocation, confounding factors, intervention administration, outcome measurement (time points, method, and reliability), participant retention, and statistical conclusion validity, which we report in the Results section.

These appraisals informed consideration of the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework [37], which was prespecified in the protocol to assess certainty of evidence across studies. However, application of GRADE across studies requires sufficient comparability in outcome definitions, measures, and reporting [38], and this was not present within any outcome domain in this review. We therefore did not undertake a formal assessment of evidence certainty or reporting bias. Further, although studies could be grouped broadly by outcome domain, substantial heterogeneity remained in participant populations and settings, intervention content and delivery, tailoring mechanisms, comparator conditions, and outcome assessment. We therefore also did not conduct a meta-analysis because we did not deem the included studies sufficiently similar to support a meaningful pooled estimate. Findings were instead synthesized narratively and organized according to key JITAI tailoring components and outcome domains, including intervention aims, tailoring variables, decision rules, and decision points. All included studies were considered in the narrative synthesis.

To ground implications of this review in lived experience, we directly engaged 12 AYAs, including health activists from the World Health Organization European Region’s Youth4Health network, a platform for AYA participation in health policymaking. These AYAs, including young entrepreneurs, medical students, and policy advocates, provided critical perspectives that informed study interpretations and implications. Seven AYAs were directly involved as coauthors in screening studies, extracting data, and conducting critical appraisals, including the lead author. AYA perspectives were further elicited through a dedicated virtual consultative session with AYAs who were not part of the coauthorship team. During this session, 5 AYAs reviewed the data extraction tables from the review and engaged in semistructured critical dialogue regarding intervention acceptability, design relevance, and the alignment of findings with lived adolescent experiences. These discussions were documented via meeting minutes and synthesized by KWG and MA to identify recurring interpretive themes. Given several of the AYA participants’ roles as coauthors in the review alongside the limited sample, these AYA contributions are reported in the Discussion section as explicitly interpretive observations to contextualize preliminary recommendations for future work, rather than as additional empirical data derived from the included studies.

A total of 3626 unique records were identified via the literature search and screened in the title and abstract phase (Figure 1). Of these, 3374 records were excluded as they were clearly focused on adult populations or not intervention studies, such as studies collecting ecological momentary assessments only. Following this exclusion, a total of 252 full-text papers were assessed for relevance. Papers that were excluded at the full-text stage did not develop mobile interventions that were tailored to dynamic adolescent contexts, for example, those offering only prescheduled digital health support without ongoing adaptation, and therefore did not meet the definition of a JITAI [16]. This resulted in a total of 69 studies (representing 61 unique interventions) developing JITAIs for AYAs included in the findings synthesis that follows.

The risk of bias and critical appraisal matrix (Figure 2) shows that every study had at least one methodological concern, indicating a moderate to high risk of bias present across the evidence base. We organize these by study design (randomized controlled trial [RCT], noncontrolled experimental studies, and qualitative studies), as critical appraisal is conducted based on the specific design of a study [36]. Importantly, many “unclear” ratings are likely attributable to incomplete reporting rather than the absence of methodological safeguards, reflecting variability in reporting practices and the lack of standardized conventions in this field. These issues significantly reduce confidence in reports of effectiveness and complicate comparisons across studies.

Intervention administration was the most prominent limitation unique to RCTs. Across studies, reporting was limited on whether treatment groups were treated identically apart from the intervention of interest. Importantly, intervention administration is inherently difficult to deliver in a blinded and standardized manner in JITAIs because these interventions are designed to deliver tailored and timely support based on momentary context, vulnerability, and receptivity, rather than static content delivered uniformly across participants [16]. As a result, intervention groups often received more personalized prompts, summaries, feedback, adaptive logic, or supportive features than comparator groups, making performance bias difficult to avoid.

Meanwhile, noncontrolled studies frequently showed weaknesses in selection procedures, as comparison groups were often absent and allocation methods lacked detailed justification. This limits confidence that observed effects can be attributed to the JITAIs being evaluated rather than to pre-existing participant differences or co-occurring influences. Again, this limitation may be especially important for JITAIs, because their effects are hypothesized to depend on timing, context, and adaptive delivery, making it necessary to isolate the contribution of specific intervention components [16]. Further, both randomized and noncontrolled studies in this study demonstrated consistent weaknesses in outcome measurement reliability, participant retention, and statistical conclusion validity. Few studies provided sufficient study-specific evidence on measurement reliability, and loss to follow-up was often incompletely reported. Further, comprehensive details on statistical power and power assumptions were frequently absent or reported with caveats, consistent with the predominance of pilot samples across the included studies.

Within qualitative studies, the most consistent limitation was researcher positioning and reflexivity, which was rarely reported or addressed. This was reflected in the limited reporting of whether the researcher’s cultural or theoretical orientation was declared. Further, the potential influence of the researcher on study participants, or vice versa, was rarely acknowledged.

Table 2 presents a domain-level synthesis of the theoretical frameworks that informed (distinct) intervention designs, alongside the primary intervention options, tailoring variables, and decision points. Note that counts overlap as studies applied multiple components simultaneously. Across all JITAIs targeting AYA health, researchers commonly used tailored digital messages, which were delivered via text or app-based messaging tools, while leveraging diverse theoretical frameworks to inform intervention content.

In AYA JITAIs targeting substance use, motivational approaches were most frequently applied (n=8, 33.3%), informing intervention components on self-reflection and dynamic goal-setting [43,45,47,50,52,53,57,59,61]. Cognitive-behavioral approaches were also frequently applied (n=6, 25%), such as informing coping prompts for managing substance cravings and related emotional distress that may trigger use [44,52,54,55,105]. Several JITAIs for AYA substance use also adopted the transtheoretical model (n=4, 16.7%), which guided the tailoring of feedback specific to an individual’s stage of behavior change [40,53,56,59]. Among JITAIs focused on AYA mental health, most (n=8, 34.8%) also leveraged cognitive behavioral approaches to inform intervention content around cognitive restructuring and stress-related coping strategies [65,70,75,76,78,85,86]. JITAIs for AYA mental health also consisted of supportive messages based on emotion regulation [67,84,85], motivational interviewing [67,76,86], mindfulness-based strategies [71,77,85], and transdiagnostic approaches [69,70,72] during heightened stress events (n=3, 13% for each respective approach). Similarly, during individually challenging moments, compassion-focused JITAIs (n=2, 8.7% within this domain) delivered content aimed at promoting emotional resilience and self-kindness [64,66,69].

For JITAIs promoting AYA physical health, self-regulation and control-based approaches were most frequently adopted (n=6, 42.9%) to tailor motivational prompts that emphasize autonomy and self-efficacy to encourage sustained health behaviors [89,91,95-97]. Around 21.4% (n=3) of JITAIs for AYA physical health also adopted the Behavior Change Wheel to guide structured prompts and reinforcement strategies, such as reminders to reduce sedentary behavior and promote physical activity, including supplementation with gamified reward systems [94,95,97]. Among JITAIs focused specifically on managing physical symptoms of chronic health conditions, frameworks such as self-regulation theory were used to structure feedback mechanisms to address adherence barriers, such as for asthma management [102].

There is an overall trend of basing adaptive content on theories related to promoting AYAs’ levels of motivation and self-efficacy, cognitive and behavioral change states, although the specific application of these principles varies by context. For example, cognitive-behavioral approaches can be broadly applied to aid in managing cravings in substance use in addition to cognitive restructuring in mental health, while also supporting adherence behaviors in chronic pain management. Behavioral intent and goal-setting also inform intervention content, with JITAIs commonly adapting their support based on AYA-defined goals and progress [63,89,91,96,97,103]. This significant variability underscores the need to understand how the adaptive selection of these dynamic interventions, such as timing, content, and delivery, align with ongoing AYA contexts, which is the focus of the next section.

Tailoring mechanisms within AYA JITAIs are primarily informed by behavioral, psychological, and other contextual data (Table 2). Further, the tailoring variables were aligned with the proximal and distal outcomes of the interventions. For example, behavioral data, such as self-reported nutritional consumption and physical activity levels, were used to prompt personalized feedback and inform the content of messages [91,95,96]. Meanwhile, psychological variables, such as fluctuations in mood and self-efficacy, informed JITAIs aimed at improving AYAs’ emotional regulation and reducing negative affect [65,67,69,70,76]. Across these approaches, contextual awareness is emphasized, with JITAIs tailoring content based on individual and environmental factors.

Among AYA JITAIs addressing substance use and mental health, most tailoring variables rely on self-reported data, such as mood, stress, and alcohol consumption. Substance use interventions predominantly tailored content to self-reports of substance use (eg, frequency, quantity, cravings, and triggers; n=16, 66.7%); additional tailoring drew upon motivational states (n=15, 62.5%) as well as high-risk social or situational contexts (eg, location, time of week, and social setting; n=11, 45.8%). One study collected electrodermal activity to determine high-risk stress states requiring intervention to prevent drinking [52]. Mental health–focused JITAIs are tailored primarily to self-reported affective states (n=15, 65.2%), such as mood, stress, and negative thinking. Some studies also tailored support based on self-reported self-efficacy and coping needs (n=5, 21.7%, respectively). Passive collection included geolocation (n=4, 17.4%), device usage (n=2, 8.7%), and electrodermal activity in one study to determine states of stress [52].

In contrast to JITAIs for AYA substance use or mental health, over half of JITAIs for AYA physical health tailored intervention delivery based on passive activity data. These include tailoring based on step counts (n=4, 28.6%) and sedentary time (n=4, 28.6%), including via sensing data using Bluetooth geofencing (n=3, 21.4%). This reflects a trend toward real-time, event-triggered decision points in JITAIs, characterized by interventions that are activated based on immediate sensor data, such as detecting physical inactivity [92,93,97]. Notwithstanding, self-reported data were also frequently leveraged in JITAIs for AYA physical health, including for monitoring behavior (n=10, 71.4%), chronic disease symptoms (n=5, 35.7%), and goal progress (n=7, 50%).

In terms of decision points, JITAI delivery is characterized by the use of daily or multiple times daily surveys to collect these tailoring variables [89,96,99-102]. These decision points enabled AYAs to prompt immediate interventions adapted based on the self-reported data. Across all AYA JITAIs, these mechanisms can also be broadly characterized as following threshold-based triggers, where responses are activated if predefined conditions are met, such as reaching a specific stress level or failing to meet activity goals [43,52,54,55,64,65,67,69,71,79,97,99,101]. This approach allowed for adaptive interventions to be activated after consistent patterns of inactivity, poor mental health, or unmet goals. Such mechanisms were further enabled through the incorporation of high-frequency, scheduled prompts, including multiple daily check-ins or morning and evening notifications [48,50,68,71,79,96,99,102,105].

Some studies also allowed for AYA-initiated check-ins to promote flexibility in need-based and real-time support [40,52,58,68,69,74,86]. Meanwhile, in studies that used behavioral monitoring, interventions were characterized by fixed-schedule decision points, which were based on predefined intervals [53,65,79,81,89,96,100,102,105]. Studies also tailored intervention schedules based on participant-defined or behaviorally relevant times, such as following peak drinking days [43,47,105]. Time of day and spatial cues have also been leveraged to refine the timing and delivery of interventions, with methods including geofencing used to automatically identify high-risk locations and detect sedentary behavior to prompt tailored notifications [46,74,97,105,106].

Since 2024, JITAIs for AYAs have leveraged artificial intelligence to enhance the adaptability of interventions. One study developed a JITAI to prevent smartphone overuse, introducing a friction-based intervention to disrupt access to blacklisted apps when overuse was detected. The authors used a machine learning model, which was updated daily with passive sensing data such as app usage, activity, social context, location, and temporal patterns to predict overuse, triggering interventions upon app launches and task-switching moments [87]. Another recent study leveraged large language models (LLMs) for dynamic adaptability, also with the goal of reducing problematic smartphone use [88]. Similarly, when AYA users opened blacklisted apps, the LLM generated persuasive messages using 4 communication strategies of understanding, comforting, evoking, and scaffolding. LLM messages were also aligned with tailoring variables, which included real-time usage, context, mental states, and user-defined goals related to app use.

A critical gap emerges in the lack of detailed reporting on AYA engagement across studies, limiting formal analysis with the tailoring mechanisms identified in prior sections. AYA’s engagement in the development of JITAI components demonstrates considerable variation in both depth and rigor. Certain studies integrated the participants directly into the design process [57,59,95,102], while others consulted AYAs with demographic profiles similar to the intended intervention audience [39,77,83]. Such engagement is frequently framed as a strategy to enhance intervention relevance and acceptability by involving adolescents in consulting on the relevance of content. However, the extent of AYA participation in these studies is also inconsistently reported, particularly within RCTs, which often prioritize the evaluation of predeveloped interventions over documentation of co-design processes. In contrast, qualitative studies focused on describing processes of adolescent engagement in JITAI component development explicitly.

While unreported involvement in these interventions is also possible, the absence of such documentation limits our ability to assess their impact on JITAI tailoring mechanisms. Moreover, among studies claiming AYA involvement, the depth of engagement was frequently superficial. It was unclear whether AYAs had substantive influence over core design decisions regarding decision rules, except for among the small sample of qualitative studies. For example, one study engaged AYAs exiting foster care in co-designing a JITAI for substance use reduction [53]. Through focus groups with these AYAs, participants provided iterative feedback on intervention structure, language, message frequency, and content relevance, which led to key modifications in JITAI content, including adjustments to message tone, content specificity, and delivery timing. Similarly, one study involved Dutch lower-educated AYAs in the iterative development of JITAI content, including informing modifications to text clarity, visual elements, navigation, and notification settings [61]. Another study involved ethnic minority adolescent females in developing a Build Your Own Theme Song app [75]. Adolescent females contributed to JITAI design by creating personal theme songs through rewriting and recording lyrics based on their experiences. Subsequently, the app would prompt adolescents to engage with their theme song during moments of negative thinking, providing a personalized coping mechanism based on their own input.

Ethical considerations were rarely addressed in detail for specific intervention components or data collection mechanisms. A minority described explicit strategies that were primarily limited to study design, such as providing devices or data plans [44,52,85], using data encryption [99], or advising participants on secure message deletion [65]. However, reporting on the ethical implications of tailoring mechanisms was notably absent. No studies discussed mechanisms for data sovereignty or the granularity of consent regarding how distinct personal data might inform algorithmic and intervention decision-making. Despite the increasing use of passive sensing and machine learning [87], there was a lack of transparency regarding how these automated decisions were explained to adolescent users. This obscures potential risks related to privacy and autonomy, highlighting an urgent need for standardized ethical reporting guidelines in AYA digital health.

This systematic review identified key patterns in how JITAIs for AYAs use dynamic tailoring mechanisms, decision rules, and adaptive data sources to personalize health interventions for improving substance use, mental health, and physical health (including chronic disease) outcomes. Most JITAIs relied on self-reported data, with growing trends toward integrating passive and sensor-based data, such as geofencing, app usage, and physiological monitoring [65,74,85,87,89]. Decision rules were frequently based on threshold triggers or scheduled intervals, though emerging studies have begun to explore LLMs and machine learning models for real-time, adaptive interventions [87,88].

The dominance of self-report behavioral data in JITAIs also suggests limited consideration of broader contextual and social determinants of health in their tailoring mechanisms, even when studies were piloted specifically with underserved communities [49,60,61,75,85,89]. Moreover, while our risk of bias appraisal assessed participant retention across studies, JITAI-specific data missingness also includes ecological momentary assessment nonresponse, sensor data loss, or selective absence of data at clinically important moments [107,108], which likely affects outcome measurement quality as well as intervention delivery. It is crucial for future work to investigate the incorporation of both social and missingness factors in JITAI tailoring to enhance contextual relevance and better address barriers to equitable health support among diverse individuals [27,109,110].

Further, JITAI mechanisms were highly heterogeneous, and confidence in the evidence base was constrained by inconsistent reporting and heterogeneous study designs. Further, while specific outcomes have yielded positive effects, most outcomes exhibit variable or limited findings [44,46,80,90,91,105]. Taken together, these patterns suggest the need for further investigation into the methodological determinants of intervention success, particularly regarding how JITAI and study design influence observed effectiveness.

It is crucial to interpret current observations of effectiveness in this review, including limited effects, with significant caution. The majority of current evidence for AYA JITAIs comes from small, pilot studies. Comparison groups were often absent or weak, and attrition or response burden was common. Several authors also explicitly noted that observed improvements could not be disentangled from self-monitoring effects, baseline reporting, treatment as usual, in-person counseling, or other concurrent influences [50,55,64]. Meanwhile, while a substantial number of studies were randomized, the risk of bias assessment highlighted recurrent challenges such as limited blinding, unclear allocation concealment, and nonidentical intervention administration. Heterogeneity in comparators also reduces confidence that the current observed effects can be attributed solely to the JITAIs being tested [111].

Given these limitations, the literature may benefit less from further RCTs but more from optimization-oriented trials, particularly microrandomized trials (MRTs), which can evaluate which intervention components work, at what moments, and for what specific AYAs’ daily life contexts. MRTs, in contrast to conventional RCTs, randomize intervention options at multiple and repeated decision points to help precisely determine the effectiveness of individual adaptive components across diverse individual contexts [112]. Such an approach can enable “a holistic comprehension of intervention dynamics over various timescales” [109]. We posit that even randomized JITAI study designs may be unable to sufficiently detect within-person variability or isolate momentary effects over time and may therefore underestimate or misrepresent the effectiveness of JITAI. To deliver precise JITAIs to AYAs, it may be important to assess diverse contextual factors that may influence the effectiveness of different adaptive components, as this can help understand “when and in what contexts a particular intervention option should be delivered”—a key differentiator of JITAIs from other digital health interventions [112]. In summary, MRTs can more precisely enable the measurement of the specific tailoring variables, decision points, and intervention options that are optimal for different individuals and health outcomes [109]. MRTs can also help to estimate proximal effects even under variable engagement and contextual missingness.

Further, for qualitative studies, the risk of bias assessment highlighted that many authors did not sufficiently consider researcher positioning and reflexivity [113]. AYA’s participation in JITAI development was often limited to feedback on usability or content refinement rather than shared involvement in intervention design. We also observed inconsistent considerations of privacy, consent, and ethical safeguards in AYA JITAI studies. These findings highlight critical gaps in ensuring that JITAIs align meaningfully with AYA contexts and needs alongside critical ethical standards [114].

This review is the first to focus on JITAIs across a wide range of health conditions within a specific developmental period, highlighting the unique needs and intervention contexts for adolescents, who face distinct physical, mental, and social health challenges that differ from those of children or adults. A key strength of this review is the meaningful involvement of AYAs in screening, data extraction, and critical interpretation, which enhances the relevance and actionability of the findings and future recommendations. The current synthesis of common theoretical frameworks and tailoring mechanisms also represents an important first step toward a more coherent body of evidence for the application of JITAIs in AYA health. However, due to substantial heterogeneity in intervention designs, high risk of bias (particularly with outcome reliability and statistical validity), and inconsistent outcome effects observed, meta-analysis was not feasible. Additionally, despite a broad search strategy, some relevant studies may have been inadvertently missed, as most included interventions did not explicitly describe their mechanisms as just-in-time or adaptive.

This review provides the first comprehensive synthesis of how JITAI mechanisms are being designed and studied for AYAs across multiple health domains, while also integrating AYA perspectives into the interpretation of findings and future directions. Beyond showing where JITAIs are being applied, it highlights important limitations in the current evidence base, including heavy reliance on self-reported behavioral data and limited attention to broader social determinants in tailoring mechanisms. The review also highlights substantial methodological heterogeneity among studies evaluating similar outcomes, as well as insufficient ethical transparency, including limited participatory engagement to ensure intervention designs are grounded in real-life AYA contexts. Our findings suggest that progress in AYA precision health may depend not only on developing more adaptive interventions but on measuring and designing them in ways that are more transparent and participatory. Limitations of traditional study designs, including measurement of JITAI administration in RCTs, point to the need for more robust designs such as MRTs as a key methodological next step for identifying which adaptive components work, at what moments, and in which daily life contexts. Based on AYA coauthor and consultant input, we further outline preliminary recommendations for future research to help ensure JITAIs better align precision with AYA agencies, including through co-designed tailoring mechanisms, reduced notification burden, and clearer control over personal data and automated decision-making.