Journal of Medical Internet Research

The Digital Exposome: A Life Course Framework for Health in the Digital Age

2026-05-08T15:00:24-04:00

Digital technologies are reshaping human behavior, health care delivery, and population health; however, their cumulative effects across the lifespan remain underexplored. This viewpoint argues that exposures arising from interactions with digital technologies should be formally integrated into exposome science as a distinct, measurable component of the human environment. Our aims are to (1) redefine the digital component of the exposome (the digital exposome) within the broader exposome framework, (2) examine its life course implications for health and equity, and (3) outline a research and policy agenda to enable its systematic measurement and integration into clinical and public health practice. Digital technology–related exposures can confer benefits such as enhanced health monitoring, personalized interventions, improved access to care, and the promotion of healthy behaviors. However, they may also introduce potential risks, including mental health challenges, cognitive and circadian disruptions, sedentary lifestyles, exposure to misinformation, and widening inequities among vulnerable populations. Despite their ubiquity, digital technology–related exposures remain poorly integrated into clinical medicine, epidemiology, or public and global health policies. Drawing on interdisciplinary evidence from exposure science, epidemiology, and digital phenotyping research, we propose a refined conceptual definition of the digital exposome grounded in the classical exposome domains. We propose redefining the digital exposome as the full spectrum of exposures resulting from interactions or proximities with digital technologies and their combined influence on health across the lifespan. This framework conceptualizes digital technology–related exposures as a dynamic set of environmental influences operating through sociotechnical, behavioral, and biological pathways over the life course. To operationalize this framework, we discuss practical approaches using validated behavioral instruments, objective device use logs, ecological momentary assessments, smartphone-based digital phenotyping, and wearable sensing technologies. Systematic measurement, large-scale longitudinal studies, and harmonized exposure metrics are needed to characterize the cumulative health impacts of digital environments more accurately. Emerging tools such as digital markers or biomarkers and digital phenotypes offer promising opportunities to link real-world technology use with physiological and biological outcomes, thereby supporting precision medicine and population health strategies. Ethical governance, privacy safeguards, and equity considerations must be embedded from the start, drawing on emerging exposomethics frameworks. Recognizing the digital exposome as a modifiable determinant of health offers a foundation for evidence-based guidance, prevention strategies, and policy interventions suited to increasingly digital societies. By integrating digital technology–related exposures into exposome science, clinical practice, and public health research, this viewpoint seeks to foster interdisciplinary dialogue, guide future empirical work, and support the development of safer and more equitable digital environments across the lifespan.

The Actionable Innovation Day Approach: Participatory Model for Advancing Critical Care Innovation

2026-05-08T14:45:13-04:00

Health care innovation is essential for improving patient outcomes, enhancing system efficiency, and preparing for future challenges; however, meaningful progress is often hindered by entrenched barriers such as resistance to change, fragmented interdisciplinary collaboration, and constrained financial and human resources. These persistent obstacles make it difficult for health care systems to translate creative ideas into sustainable, real-world improvements, underscoring the need for structured approaches that support collaboration and reduce implementation friction. To address these challenges, we developed the Actionable Innovation Day (AID) approach, a structured, participatory model designed to generate consensus-based, low-cost recommendations that are feasible for system improvement. The first regional AID event in Eastern Ontario gathered 57 multidisciplinary participants, including clinicians, administrators, patient partners, and industry leaders, for a full-day series of presentations, facilitated discussions, and targeted breakout sessions focused on critical care. Through guided deliberation and collaborative analysis, participants synthesized diverse perspectives into a prioritized set of improvement opportunities. The process yielded 28 actionable recommendations across 4 domains: health care innovation, regionalized care, critical care practices, and the use of artificial intelligence. A postevent survey (86% response rate) showed strong agreement, with 23 recommendations rated above 4 on a 5-point scale. The highest-ranked proposals emphasized the value of strengthening research-industry-clinical partnerships, integrating families more intentionally into intensive care unit rehabilitation and recovery processes, and implementing centralized regional coordination to optimize critical care capacity. Together, these findings illustrate not only the feasibility of the AID model but also the AID model’s ability to surface strategic, context-appropriate solutions that resonate across stakeholder groups. The AID process offers a scalable and adaptable template for advancing health care innovation through collaborative, real-world problem-solving. While this initial event focused on critical care, the underlying principles of structured engagement, iterative consensus building, and interdisciplinary co-design are broadly applicable to many sectors of health care. We encourage institutions, regional networks, and health system leaders to adopt and tailor the AID framework to their own local priorities, recognizing that inclusive innovation processes can accelerate system improvement even in resource-limited settings. Ultimately, the AID approach serves as both a methodology and a call to action: by empowering teams to collectively identify, refine, and champion actionable ideas, health care organizations can build the capacity and culture necessary to drive meaningful and sustained innovation across diverse clinical and operational domains.

Efficacy of Technology-Based Interventions on the Reduction of Loneliness: Systematic Review and Meta-Analysis

2026-05-08T13:00:23-04:00

Background: Loneliness is a widespread public health concern linked to increased risks of health problems. As populations age, the demand for effective interventions to mitigate loneliness continues to grow. Objective: This meta-analysis aimed to examine the effectiveness of technology-based interventions in reducing loneliness. Methods: A systematic literature search was conducted in Web of Science, PsycInfo, PubMed, Scopus, Google Scholar, Embase, and the Cochrane Library in August 2024. We included randomized controlled trials that examined the effectiveness of technology-based interventions compared with any control group in reducing loneliness across all age groups. Nonrandomized studies, qualitative research, and studies lacking sufficient statistical data for effect size calculation were excluded. After screening 1089 records, 7 studies involving 580 participants met the inclusion criteria. Data were extracted by 3 independent reviewers, with discrepancies resolved by a fourth reviewer. The risk of bias was assessed using the Cochrane risk-of-bias tool. A random-effects model was used to synthesize effect estimates, with standardized mean differences as the primary effect size metric. Heterogeneity was assessed using the Q statistic and ² index, and a prediction interval was calculated to estimate the expected range of true effects. Results: We found a small and statistically nonsignificant overall effect of technology-based interventions on loneliness (pooled standardized mean difference=–0.21, 95% CI –0.59 to 0.17; 95% prediction interval –1.14 to 0.63). Substantial between-study variability was present (²=57%; τ²=0.08; =0.28), and the prediction interval indicated that true effects in future studies may range from substantial reductions to moderate increases in loneliness. Differences between intervention types could not be examined due to the limited number of eligible studies. The Egger test showed significant funnel plot asymmetry (=–5.47; =.003). However, since the trim-and-fill method identified no missing studies, the asymmetry is unlikely to be fully explained by publication bias. Conclusions: Unlike previous reviews, which focused primarily on older populations or included nonrandomized studies, this meta-analysis provides a rigorous synthesis of only randomized controlled trials across the lifespan. Our findings do not provide evidence for a reliable reduction in loneliness following technology-based interventions. Moderate heterogeneity indicates that effects differ across studies, and the limited number of eligible trials prevented the analysis of potential moderators. By incorporating robust prediction intervals and a broader age demographic, this study offers a more comprehensive view of the variability of intervention outcomes compared with existing literature. The wide prediction interval suggests that intervention effects may vary across settings. In practice, technology-based approaches should be viewed as supportive tools that require careful tailoring rather than universal solutions to loneliness. Trial Registration: PROSPERO CRD42024505117; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024505117

Governing Patient-Facing AI-Generated Video in Digital Health: A Risk-and-Ethics Matrix for Deployment, Monitoring, and Change Control

2026-05-08T13:00:23-04:00

In this Viewpoint, we argue that patient-facing high-fidelity artificial intelligence (AI)–generated video requires governance that is operational, life cycle based, and embedded in existing institutional review pathways rather than limited to predeployment checks alone. Patient-facing high-fidelity AI-generated video—synthetic or substantially AI-mediated video that presents realistic human likeness, voices, or clinical communication cues—is rapidly entering patient education and clinical communication. We propose a risk-and-ethics matrix that combines residual clinical risk (likelihood × severity after mitigations) with an ethical alignment score that operationalizes autonomy, beneficence, nonmaleficence, and justice to yield actionable dispositions (encourage, permit with oversight, restrict or redesign, or prohibit). The framework links each disposition to dossier-based review, minimum controls, and postdeployment monitoring triggers—focused on measurable outcomes (eg, comprehension, content-attributable follow-up burden, incidents and complaints, and equity gaps) as well as provenance and change control—to support auditable, revisitable decisions over the system life cycle.

Barriers and Facilitators to Patient Acceptance of Artificial Intelligence in Health Care: Systematic Review

2026-05-08T10:00:22-04:00

Background: Artificial intelligence (AI) in the domain of health care is increasing in prominence. Acceptance is an indispensable prerequisite for the widespread implementation of AI. Objective: This study aimed to explore barriers and facilitators influencing patients’ acceptance of AI. Methods: We conducted a systematic review following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Nine databases, including PubMed, Web of Science, and Embase, were comprehensively searched from inception to December 23, 2025. We included qualitative, quantitative, and mixed methods studies investigating adult patients’ attitudes toward medical AI. Two researchers independently screened records, extracted data, and appraised methodological quality using the Mixed Methods Appraisal Tool. Following the Joanna Briggs Institute convergent integrated approach, data synthesis was guided by integrating the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) and the Theoretical Domains Framework (TDF). Factors were mapped to behavior change techniques (BCTs) and evaluated for practical feasibility using the Affordability, Practicability, Effectiveness and cost-effectiveness, Acceptability, Side-effects/safety, and Equity criteria. Results: A total of 61 studies met the inclusion criteria out of 7452 search results. Study designs included qualitative (n=20), quantitative (n=35), and mixed methods (n=6). Performance and effort expectancies were the primary determinants of acceptance. Major barriers comprised perceived operational complexity, lack of algorithmic trust, reduced interpersonal interaction, privacy vulnerabilities, and high costs. Facilitators included transparent data governance, interpretability of AI decisions, improved clinician-patient communication, and human-centered design. Education level and disease severity emerged as key moderating variables. Through UTAUT2-TDF mapping, we identified 25 distinct BCTs (6 high, 14 medium, and 5 low feasibility) and formulated 40 actionable intervention strategies. Conclusions: This study innovatively integrates the UTAUT2 and TDF frameworks to evaluate patient acceptance of medical AI. Unlike existing reviews that predominantly evaluate isolated psychosocial factors or purely technical attributes, this transtheoretical approach differentiates itself by merging technology adoption mechanisms directly with behavioral drivers. Consequently, it contributes to the field by systematically identifying multilevel factors influencing acceptance, including performance expectancy, effort expectancy, and ethical security, and translating these into 40 actionable BCTs. In real-world clinical practice, these findings provide a feasible, prioritized blueprint for clinicians and administrators to design patient-centered interventions, enhancing the clinical integration and long-term effectiveness of medical AI. Trial Registration: PROSPERO CRD42024598884; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024598884

Efficacy, User Engagement, and Acceptability of Cognitive Behavioral Therapy–Oriented Psychological Chatbots for Adults With Depressive and/or Anxiety Symptoms: Systematic Review and Meta-Analysis of Randomized Controlled Trials

2026-05-08T10:00:22-04:00

Background: Cognitive behavioral therapy (CBT) is the most examined psychotherapy for depression and anxiety, but delivery faces significant barriers such as limited access, cost, and time constraints. CBT-oriented psychological chatbots offer a promising means of addressing these challenges. Yet, their overall efficacy, user engagement, and acceptability have not been systematically synthesized. Objective: This study aimed to evaluate the efficacy, user engagement, and acceptability of CBT-oriented chatbots for adults with depressive and/or anxiety symptoms. Methods: A systematic search of 9 databases, including PubMed, Cochrane Central Register of Controlled Trials, Embase, Web of Science, PsycINFO, CINAHL, China National Knowledge Infrastructure, WanFang, and VIP Databases, was conducted from inception to February 2026. Eligibility criteria included randomized controlled trials comparing CBT-oriented chatbots with control groups in adults with depressive and/or anxiety symptoms. Risk of bias (ROB) was assessed using the Cochrane ROB tool. Random-effects meta-analyses (Hartung-Knapp-Sidik-Jonkman adjustment) calculated pooled effect sizes (Hedges ), 95% CIs, and 95% prediction intervals (PIs). Heterogeneity was evaluated using the ² statistic, and Galbraith plots were used to identify outliers for subsequent sensitivity analyses. Subgroup and meta-regression analyses examined potential moderators. The certainty of evidence was evaluated using the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) approach. Data on user engagement and acceptability were extracted and synthesized using narrative and quantitative methods where available. Results: Twenty-nine eligible randomized controlled trials were included. CBT-oriented psychological chatbots produced a moderate reduction in depressive symptoms at postintervention (=−0.55, 95% CI −0.70 to −0.40, 95% PI −1.23 to 0.13) and a small reduction in anxiety symptoms (=−0.26, 95% CI −0.37 to −0.14, 95% PI −0.67 to 0.15). At follow-up, effects were small for depression (=−0.32, 95% CI −0.55 to −0.09, 95% PI −0.93 to 0.29) and nonsignificant for anxiety (=−0.19, 95% CI −0.43 to 0.04, 95% PI −0.84 to 0.46). Subgroup and meta-regression analyses revealed that anxiety outcomes were significantly moderated by clinical profiles—showing distinct advantages for comorbid symptoms—and the proportion of female participants. The CBT-oriented chatbots received an adequate level of engagement that complied with digital intervention standards. Although user satisfaction ratings were generally favorable, technical limitations and repetitive interaction patterns remain to be addressed to enhance overall acceptability. Regarding the limitations of evidence, the overall certainty was rated as very low to low, predominantly driven by high ROB and substantial heterogeneity. Conclusions: This study innovatively isolates CBT-oriented chatbots from broader digital interventions, providing a precise, methodology-driven evaluation of theoretically grounded therapeutics. This review brings critical evidence to the field that these tools yield significant short-term relief, particularly for comorbid anxiety profiles. In the real world, CBT chatbots offer profound potential as scalable, low-barrier first-line tools. To sustain engagement, future developments must evolve from rigid rule-based scripts toward adaptive, large language model–driven architectures while ensuring clinical safety. Trial Registration: PROSPERO CRD42024615506; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024615506

Diabetes Technologies in Ultra-Endurance Type 1 Diabetes: Qualitative Study

2026-05-08T09:30:13-04:00

Background: Diabetes technologies—including continuous glucose monitoring (CGM), insulin pumps, and hybrid closed-loop systems—have profoundly transformed self-management in type 1 diabetes (T1D). While these technologies offer improved glycemic control and safety, their use in ultraendurance sports introduces specific cognitive, material, and organizational challenges that remain underexplored in digital health research. Objective: This study aimed to explore how adults living with T1D experience and use diabetes technologies in ultraendurance sports, with particular attention to tensions between autonomy, mental load, and vulnerability. Methods: We conducted semistructured interviews with 13 French-speaking adults with T1D who had completed at least one marathon or ultra-endurance event within the last 5 years and used ≥1 diabetes technology (CGM, pump, or hybrid closed loop). We adopted constructivist grounded theory (Charmaz), using iterative cycles of line-by-line and focused coding, constant comparison, and memo-writing to build and refine analytic categories. Sampling combined purposive strategies through associations and online communities with theoretical orientation (additional participants sought to elaborate emergent categories). Data collection ceased upon theoretical sufficiency, when further interviews no longer yielded substantively new insights for core categories. Two patient partners contributed to question framing, interim sense-checking, and manuscript review. Reporting followed the COREQ (Consolidated Criteria for Reporting Qualitative Research) checklist. Results: Five interrelated categories described how athletes negotiated technology in practice: (1) From episodic control to continuous anticipation (reframing glucose management through real-time visibility); (2) Gains in safety and performance (perceived benefits and expanded possibilities); (3) Redistributed mental work (hyper-vigilance, logistics, device management); (4) Keeping things working when they break (fragility in extreme conditions, redundancy, improvisation, and experiential expertise); and (5) Making diabetes visible (technologies mediating identity, solidarity, and stigma). Across categories, participants articulated a tension between optimization-oriented performance and a user-constructed robustness—the capacity to maintain function under uncertainty through redundancy and adaptive know-how. Conclusions: In ultraendurance contexts, diabetes technologies act as both enablers and obligations: they open participation while shifting and sometimes intensifying cognitive and organizational work. A grounded account centered on robustness-in-use highlights practical implications for clinicians (pre-event routines, redundancy planning), designers (context-aware algorithms; improved physical durability), and policy makers (equitable access and exercise-specific education). These findings underscore the value of constructivist, practice-oriented inquiry to inform digital health tool design and support for people living with chronic illness.

Thematic Mapping and Evolution of Social Media Mining in Health Research: Hybrid Bibliometric Synthesis

2026-05-08T09:00:24-04:00

Background: Social media platforms offer extensive data, as they are widely used globally. Social media mining (SMM) enables real-time monitoring of user-reported health information and serves as a supplement to traditional health data analytics. However, the rapid proliferation of literature has produced fragmentation, and a comprehensive knowledge map regarding SMM is lacking. Further, existing bibliometric reviews in health fields are easily undermined by synonym fragmentation and parameter settings, reducing their robustness. Thus, a more robust, reproducible, and decision-oriented bibliometric framework is required. Objective: This study aimed to (1) outline key thematic clusters in health-related SMM and map their dynamic evolution, and (2) methodologically demonstrate how machine learning–based bibliometric analysis can strengthen the robustness, transparency, and foresight capacity of evidence synthesis. Methods: This study designed a fully automated and reproducible bibliometric analysis of PubMed journal articles published from 2015 to 2025 (n=250) and analyzed records with both abstracts and keywords (n=189). We performed cleaning and standardization for titles, abstracts, author keywords, and MeSH terms, and carried out an exploratory descriptive analysis to obtain preliminary insights into publication patterns. Subsequently, we used SPECTER2 and PubMedBERT embeddings with keywords and abstracts to construct a hybrid similarity matrix. Then, we applied Uniform Manifold Approximation and Projection for dimensionality reduction, followed by Hierarchical Density-Based Spatial Clustering of Applications with Noise for thematic clustering, and visualized the results in a 3D strategic coordinate system (maturity, influence, and recency). We performed intercluster relationship analysis and time-slice analysis to examine thematic intersections and evolution. To ensure robustness and enhance interpretability, we implemented dual-level validation. Results: We identified 6 thematic clusters: cluster 1 (candidate incubator pool of peripheral cross-cutting topics in health-related SMM), cluster 2 (computational methods in health informatics), cluster 3 (public attitudes and sociopsychological determinants), cluster 4 (infodemiology and the COVID-19 information ecosystem), cluster 5 (health communication and public health engagement), and cluster 6 (social media analysis and network methods). Strategic 3D mapping revealed that methodological clusters (clusters 2 and 6) occupied high-maturity and high-influence positions, while application-driven clusters (clusters 3 and 4) occupied high-influence and high-recency positions, representing rapidly expanding frontiers. Clusters 1 and 5 demonstrated strong potential for further growth. Temporal slicing confirmed a trajectory moving from methodological consolidation and thematic diversification to a renewed focus on convergence and problem-solving. Validation showed strong semantic coherence and robustness of the methods and findings. Conclusions: We developed a semantic-structural hybrid bibliometric framework with dual-level validation, reducing synonym fragmentation and parameter sensitivity inherent in traditional approaches. The resulting decision-oriented knowledge map offers strategic guidance for infodemiology-informed and audience-segmented public health communication, research priority settings, and the deployment and evaluation of real-world surveillance and pharmacovigilance workflows while supporting evidence-driven and patient-centered decision-making in public health and health care.

Automated Approaches of Text Simplification of Patient Education Materials: Scoping Review

2026-05-07T17:02:57-04:00

Background: Patient education materials (PEMs) often exceed the American Medical Association’s (AMA) recommended sixth-grade reading grade level (RGL). While artificial intelligence (AI) offers potential for automated text simplification, concerns persist regarding linguistic quality, content fidelity, and the understandability of simplified PEMs by laypeople. Objective: This scoping review maps existing evidence on automated language processing technologies for simplifying PEMs for laypeople. Methods: Following the Joanna Briggs Institute (JBI) methodology and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guideline, 5 bibliographic databases (Ovid MEDLINE, Embase, CINAHL, PsycInfo, and IEEE Xplore) were systematically searched from 2019 to May 2025, supplemented by reference screening and gray literature searches. Eligible sources were peer-reviewed empirical studies published in English that examined large language models (LLMs), AI-supported writing assistants, AI-based conversational agents, or AI-supported tools designed for automatic text simplification of PEMs. Targeted outcomes included linguistic quality (ie, linguistic comprehensibility, linguistic correctness) and content fidelity (ie, factual accuracy, factual completeness) of simplified PEMs. Excluded sources comprised rule-based systems, manual text simplification, non-laypeople target groups, and technology-focused performance metrics. Results were synthesized via thematic analysis across the domains of targeted outcomes. In accordance with JBI methodology, a risk-of-bias assessment was not performed. Results: A total of 31 eligible studies met the inclusion criteria, examining various LLMs, including OpenAI’s GPT series, Gemini, Bard, Claude, Copilot, and Llama. Specifically, GPT-4.0 achieved the most consistent improvements in standardized readability metrics (eg, the Flesch-Kincaid Grade Level [FKGL]). However, achieving predefined target RGLs remained challenging across all LLMs, particularly at lower RGLs. Findings on content fidelity were inconsistent: despite high content similarity scores, content accuracy was often compromised. Conclusions: This is the first scoping review to comprehensively synthesize evidence on automated technologies for text simplification in PEMs. The review identified 2 critical validation gaps. First, no study examined the linguistic correctness (eg, grammar and typographical errors) of automatically simplified PEMs. Second, and most notably, the understandability of the simplified PEMs was assessed exclusively by experts, with no empirical evaluation involving laypeople. Although LLMs effectively reduce text complexity as measured by objective readability metrics, reliance on these formulas represents a critical limitation, as they serve merely as structural proxies. Improvements in readability do not guarantee the maintenance of content accuracy or laypeople’s understandability. Current evidence is further limited by the lack of systematic prompt quality evaluation and the predominant focus on English-language PEMs in US contexts, restricting generalizability. This review provides a foundation for future research by highlighting the need for validated evaluation frameworks that encompass layperson testing and content verification. For clinical practice, LLMs should currently serve as assistive tools, with mandatory expert review remaining essential to verify content fidelity before disseminating LLM-simplified PEMs to laypeople.

GenAI-Supported Virtual Patients in Health Care Education: Systematic Review

2026-05-07T17:00:25-04:00

Background: Generative artificial intelligence (GenAI) is enhancing virtual patient simulations in health care education by enabling dynamic, adaptive interactions, reshaping how clinical skills are taught. A synthesis of the current evidence is needed to guide implementation and future research, given the pace of technological advancement. Objective: This systematic review aims to synthesize empirical research on the design, implementation, and educational impact of GenAI-supported virtual patients in health care education. Methods: A systematic search was conducted across 5 databases (CINAHL, Medline, Embase, Scopus, and Web of Science) from their inception to March 19, 2026. Reference lists of included studies and relevant systematic reviews were also screened. Peer-reviewed studies in English that evaluated GenAI-supported virtual patients using quantitative or mixed methods were included. Two reviewers independently screened studies and extracted data. Study quality and risk of bias were assessed critically using JBI (Joanna Briggs Institute) checklists, with disagreements resolved by consensus. Results: A total of 15 studies met the inclusion criteria (total participants N=645), spanning health care disciplines, including nursing, medicine, pharmacy, radiography, and medical first-responder training. The virtual patients varied in design; input modalities included text (9 studies), voice (5 studies), or hybrid (1 study); output was text (9 studies), speech (5 studies), or both (1 study); 6 studies used 3D-embodied avatars, while 9 used nonembodied interfaces. A total of 13 studies used OpenAI GPT models (eg, ChatGPT), 1 used a fine-tuned model from a different provider, and 1 evaluated multiple model families (Claude, GPT, and open-source). Further, 6 studies used controlled experimental designs, including 3 randomized controlled trials (RCTs); the remainder were cross-sectional or prepost evaluations. Primary outcomes included user perceptions (14 studies), communication skills (4 studies), clinical reasoning (3 studies), and performance (7 studies). In controlled comparisons, GenAI-supported virtual patients consistently improved outcomes relative to control conditions: for example, enhanced clinical decision-making (RCT, n=21), ophthalmology history-taking skills (RCT, n=26), and medical history-taking performance (crossover RCT, n=20). The evidence base is characterized by brief intervention durations, a predominant reliance on single-session interactions, and a general lack of underpinning educational theory. No meta-analysis was performed due to the limited number of studies and significant heterogeneity in designs, interventions, and outcome measures. Conclusions: The evidence supports the feasibility and acceptability of GenAI-supported virtual patients, with positive learner perceptions and promising outcomes for skills development. However, critical limitations remain in emotional-behavioral complexity, simulation adaptability, and research design rigor (eg, limited use of control groups and validated instruments). The review offers educators, instructional designers, and policymakers actionable insights for integrating dynamic, artificial intelligence–driven simulations while identifying crucial gaps—such as the need for theoretical grounding, longitudinal studies, and standardized design protocols—that must be addressed for safe and effective implementation. Trial Registration: Open Science Framework (OSF) q8b5n; https://osf.io/q8b5n/files/mysz3