Anxiety-spectrum disorders are among the most common and costly mental health conditions worldwide. Data shows that about 284 million people are affected globally, leading to significant disability across all ages and economic backgrounds (Bandelow et al., 2023; Whiteford et al., 2013). While standard medication and therapy work for some, they face major hurdles. These include high dropout rates of 20–30 percent for exposure therapies, a lack of available therapists, and geographic barriers. Additionally, it is logistically difficult to provide traditional in-person exposure therapy with the steady control needed for the best results (Emmelkamp and Meyerbröker, 2021; Carl et al., 2019). Because of these issues, there is growing interest in technology-based interventions. These tools can provide evidence-based exposure therapy with more flexibility and better scaling than traditional clinics. Virtual Reality Exposure Therapy (VRET) has become the most thoroughly researched and supported option among these new methods.

VRET puts the basic principles of exposure therapy (fear activation, inhibitory learning, and the strengthening of extinction) into practice through immersive computer environments. These digital spaces allow for the presentation of feared stimuli in a way that is standardised, repeatable, and precisely controlled (Morina et al., 2015; Botella et al., 2017).

Three main concepts support how VRET works. First is presence, which is the personal feeling of “being there” in the virtual world; this acts as the main psychological bridge to trigger fear. Second is immersion, which refers to the actual technical ability of the hardware and software to create a surrounding sensory environment. Third is embodiment, or the extent to which a patient feels a virtual body is their own, a factor that is especially important for treating trauma or body image issues (Riva et al., 2019; Maples-Keller et al., 2017).

Compared to imaginal exposure, VRET has a clear clinical benefit because it does not require the patient to create vivid mental images. This is helpful for trauma survivors who may struggle with imagery or for patients who avoid mentally engaging with their fears during traditional therapy. The technology has changed significantly since it started. It has moved from the costly, lab-based CAVE systems of the 1990s to modern, affordable headsets that work wirelessly and do not need a specialized technical setup (Philippe et al., 2022; Rizzo and Koenig, 2017).

Even after 20 years of VRET development, the field still lacks a thorough quantitative map of its intellectual structure, collaboration networks, thematic changes, and publication trends for the entire 2015 to 2025 period. Past reviews have mostly focused on meta-analyses for specific disorders. Most notably (Carl et al., 2019), looked at anxiety disorders in general (Opriş et al., 2012), focused on specific phobias, and (Kothgassner et al., 2019) examined PTSD. Other works have provided narrative reviews of certain technical advances, but none have used bibliometric science mapping to describe the overall intellectual architecture or measure the structural traits of the field across all clinical areas at once.

Bibliometric analysis is a necessary addition to a systematic review because it allows for a quantitative look at a field’s structure that a qualitative summary cannot provide. This includes citation network patterns, the boundaries of thematic clusters, how concepts rise and fall over time, and international partnership trends (Donthu et al., 2021). Earlier bibliometric reviews concerning VR in clinical settings have focused on rehabilitation (Laver et al., 2025), managing pain, or specific types of disorders. However, none have fully mapped VRET as one unified field covering all clinical areas, technical aspects, and the complete 2015 to 2025 growth era. This study addresses that gap by offering both structural bibliometric mapping and a qualitative clinical summary in one combined analysis.

Research gaps in the current literature can be divided into three areas:

Theoretically, most trial designs do not properly show how hardware immersion leads to clinical results, particularly the path from feeling present in a virtual space to achieving extinction learning.

This systematic review and bibliometric analysis has four main objectives:

To evaluate the scholarly evolution of the field by quantifying growth trends, citation impact, and the distribution of research across global institutions and specialized journals;

Each of these goals is covered in the Results and Discussion sections. Objective 1 is found in Sections 3.1 through 3.4, Objective 2 is in Sections 3.5 and 3.6, Objective 3 is in Sections 4.1 through 4.4, and Objective 4 is in Section 4.5.

As shown in Figure 1, the article is structured into five cohesive sections, beginning with an Introduction that establishes the clinical background of anxiety-spectrum disorders and justifies the research objectives. This is followed by Materials and Methods, which details the database search strategy, eligibility criteria, and data extraction protocols. The Results section presents a comprehensive bibliometric overview, including publication trends, author networks, and thematic mapping. These findings are interpreted in the Discussion, which explores clinical application domains, technology integration, and therapeutic mechanisms while addressing implementation challenges and research gaps. Finally, the Conclusion synthesizes the key insights and provides a closing summary of the study’s contributions to the field.

Scopus and Web of Science were selected as the indexing databases for this study because they offer the most comprehensive coverage of peer-reviewed literature in clinical psychology, psychiatry, and computer science regarding VRET. These platforms provide the robust, structured bibliometric metadata required for science mapping and are recognized as the standard dual-database pairing for systematic bibliometric reviews (Donthu et al., 2021).

PubMed was not used as a primary database because its tools for exporting bibliometric data are limited compared to Scopus and Web of Science, and its medical content overlaps significantly with what is already in Scopus. PsycINFO was excluded because it lacks the citation network and keyword data required for Bibliometrix-based science mapping. While IEEE Xplore was considered, it was left out because its focus on electrical engineering and hardware would have brought in too much irrelevant content. Future research might consider adding PubMed and PsycINFO searches to find clinical trial registrations that these two primary databases might miss.

In terms of conceptual boundaries, this review focuses on Virtual Reality (VR) defined as fully immersive digital environments viewed through head-mounted displays or CAVE systems. Augmented Reality (AR) and Mixed Reality (MR) are included only when they are the main way the exposure is delivered. Non-immersive and screen-based virtual environments are excluded because they do not reach the level of presence required to be classified as VRET.

Conference proceedings were left out of the main analysis because the bibliometric data for these papers in Scopus and Web of Science lacks the citation stability and source metrics needed for science mapping. The authors recognize that this may mean some important contributions to human-computer interaction and VR engineering from events like CHI or IEEE VR/ISMAR were missed, which is noted as a limitation in Section 4.6.

The search across dual databases resulted in the retrieval of 1,551 articles from Scopus and 264 from Web of Science, providing a total of 1,815 records published between 2015 and 2025.

To ensure a clean dataset for quantitative bibliometric analysis, the Bibliometrix mergeDbSources function was utilized to identify and eliminate 232 overlapping documents. This process revealed a duplication rate of 12.8%, culminating in a final merged corpus of 1,583 unique articles (Table 1).

The relatively low overlap observed between these two databases highlights their complementary indexing profiles within the VRET domain. Scopus accounted for the significant majority of the records, largely due to its extensive coverage of interdisciplinary and technology-focused publications. Conversely, Web of Science served as a critical source for cross-validating the most influential core clinical literature, ensuring that the merged dataset captured both the technical and medical facets of the field.

The qualitative study selection and eligibility followed PRISMA 2020 guidelines, as shown in Figure 2 (Page et al., 2021). To be included, articles had to be peer-reviewed empirical studies published in English between 2015 and 2025 within the fields of psychology, psychiatry, neuroscience, computer science, or health sciences. They also had to focus directly on VRET as a clinical intervention. While the search terms included broader conditions like depression and psychosis to capture as much literature as possible, the final criteria required that each article specifically focus on VRET as an exposure-based treatment. Studies using VR for other types of therapy were removed during the screening process.

The quantitative bibliometric part of the study is organized as shown in Figure 3. For this process, data was gathered from all 1,583 articles deduplicated, including publication years, journals, authors, citations, and keywords. The analysis included productivity, citation trends and mapping how keywords appear together, analyzing thematic maps based on how central or dense certain topics are, and identifying trending topics by looking at median publication years. To keep the focus on significant terms, only keywords appearing at least 10 times were included in the network analysis. The study used association strength normalization to account for how often different keywords appear. A specific method called the Louvain clustering algorithm was used to group these topics naturally. In the resulting diagrams, the size of each point reflects its importance, while the thickness of the lines shows how often those topics appear together.

For the qualitative synthesis, 35 articles were chosen based on their high citation counts to ensure they represent the most influential work in the field. Because relying on citations naturally favors older papers, 5 recent high-quality studies, such as (Yang et al., 2025; Nedungadi et al., 2025; Bergsnev and Sánchez Laws, 2026; de Haart et al., 2026; Medeiros et al., 2026), were added to the discussion on technology and future trends even though they are too new to have reached the citation threshold. Totally, 40 selected studies were analyzed using a consistent framework. This covered the study design, the patient population, the specific VR hardware and software used, and the details of the therapy sessions, such as how long they lasted and the role of the therapist. It also tracked the results, effect sizes, and how long patients were followed after treatment as shown in Table 2.

The temporal analysis of the 1,583-article dataset indicates a steady and sustained expansion of VRET research between 2015 and 2025, as illustrated in Figure 4.

Annual publication volume rose from 76 articles in 2015 to 281 in 2025, marking a nearly 3.7-fold increase over the 10-year period. Output remained relatively constant between 2015 and 2018, with annual counts ranging from 75 to 100 articles, before transitioning into a phase of more rapid growth starting in 2019, when the field recorded a significant increase to 148 publications.

This trajectory is reflective of the accelerating hardware accessibility and clinical validation efforts that defined VRET research in the late 2010s. This period coincided with the broader availability of consumer-grade head-mounted displays and the increasing publication of randomized controlled trial (RCT) evidence across various anxiety disorders.

This growth trend intensified after 2021, with publications climbing from 158 in 2021 to 214 in 2024 (35.4% increase) before reaching a peak of 281 articles in 2025, the highest annual output recorded during the review period. Notably, the COVID-19 pandemic during 2020–2021 did not result in the publication disruptions observed in other clinical research areas. Instead, the consistent output of 139 articles in 2020 and 158 in 2021 suggests that the pandemic may have acted as a catalyst for VRET research. This interest was likely driven by the highlighted limitations of traditional in-person therapy and a subsequent acceleration in investment toward remote and technology-assisted mental health solutions (Horigome et al., 2020; Smith et al., 2020).

The 13.97% annual growth rate observed in the VRET literature compares favorably with broader digital health fields, where annual publication growth rates of 8%–12% have been reported for digital therapeutics generally and approximately 15% for telemedicine research, suggesting that VRET’s expansion reflects sustained field-level momentum rather than isolated publication inflation (Donthu et al., 2021). The duplication rate of 12.8% between Scopus and Web of Science is consistent with overlap rates reported in other mental health bibliometric studies and confirms that both databases contribute unique coverage to the merged corpus.

The citation metrics detailed in Table 3 highlight significant temporal shifts in research impact across the unified dataset. As expected, older publications exhibit higher mean citations per article, reflecting the extended duration available for citation accumulation. Specifically, 2017 recorded the highest mean total citations per article (71.24), followed by 2018 (57.91) and 2019 (55.72). When normalizing for differing citable lifespans using the mean citations per year metric, 2017 remained the most impactful year with an annual citation rate of 7.12 per article, followed by 2019 (6.96) and 2018 (6.43). These figures suggest that the 2017–2019 period was a particularly influential era that produced the foundational research shaping the subsequent intellectual development of the VRET field.

As expected, more recent publications show lower absolute citation counts because of their limited visibility windows; articles from 2024 averaged 6.58 total citations (2.19 per year), while 2025 articles averaged 1.67 citations. Nevertheless, the accelerating volume of publications indicates that collective knowledge production in VRET has reached its zenith within the current review period, reflecting a field that continues to expand in both scope and scholarly output.

The research landscape within the 1,583-article dataset is characterized by several key journals that maintain a sustained focus on VRET and the application of immersive technology in clinical contexts (Table 4). Cyberpsychology, Behavior, and Social Networking was identified as the leading venue with 32 articles, reflecting the field’s deep roots in the psychosocial and behavioral aspects of virtual interaction. This was followed closely by Trials (30 articles), which underscores the critical role that RCT methodology plays in building the clinical evidence base for VRET. The Journal of Medical Internet Research contributed 29 articles, further establishing digital health and technology-assisted interventions as a central pillar of the literature. Additionally, the International Journal of Environmental Research and Public Health (IJERPH) and Public Health and the Journal of Anxiety Disorders each published 26 articles, confirming that anxiety-spectrum conditions remain the primary clinical focus of the field.

The diverse range of journals highlights reflects the multidisciplinary framework of VRET research, which effectively bridges clinical psychology, psychiatry, digital health, and specialized VR technology. BMJ Open and Frontiers in Psychiatry contributed 25 articles each, demonstrating consistent engagement from the mainstream psychiatric community. Meanwhile, the Journal of Clinical Medicine (24 articles) and PLOS ONE (21 articles) indicate a broad interest in VRET that extends beyond specialist psychology circles. The Annual Review of Cybertherapy and Telemedicine, with 18 articles, has established itself as a dedicated specialized venue for research focusing on immersive therapeutic technologies.

The inclusion of foundational clinical psychology journals, such as Behaviour Research and Therapy (16 articles), BMC Psychiatry (15 articles), and Psychological Medicine (12 articles), reflects the strong theoretical anchoring of VRET within established cognitive-behavioral and experimental psychopathology traditions. Furthermore, rehabilitation-focused outlets like the Journal of Neuroengineering and Rehabilitation (15 articles) suggest significant cross-pollination between VRET and the motor rehabilitation applications of immersive tools. Technology-specialized journals, including Frontiers in Virtual Reality (14 articles) and Virtual Reality (13 articles), further illustrate the breadth of the research landscape by capturing hardware, software, and human factors dimensions alongside clinical outcome data.

Emerging interdisciplinary venues, such as JMIR Mental Health (11 articles), Current Psychiatry Reports (11 articles), and Clinical Psychology and Psychotherapy (10 articles), show the technology’s expanding footprint across both open-access digital platforms and traditional subscription-based journals. The Journal of Behavior Therapy and Experimental Psychiatry (12 articles) and the Journal of Affective Disorders (10 articles) further confirm that VRET research has achieved meaningful penetration into core behavior therapy and affective science venues, moving well beyond its origins in niche cybertherapy outlets. Collectively, the distribution of the merged dataset across 720 unique journals demonstrates the comprehensive coverage achieved by combining Scopus and Web of Science, effectively capturing the full clinical and technological spectrum of VRET research.

The analysis of author productivity within the 1,583-article dataset highlights a concentrated level of research activity among a cohort of prolific scholars who have defined the intellectual trajectory of the VRET field (Figure 5; Table 5). Giuseppe Riva emerged as the most productive author with 34 articles, reflecting a sustained research program that spans the theoretical foundations of presence and embodiment, alongside clinical applications in eating disorders, body image disturbances, and anxiety. Barbara Rothbaum followed with 31 articles, representing the field’s most consistent work on VRET for PTSD and military trauma, with a specific focus on the comparative effectiveness of VR versus traditional prolonged exposure protocols in veteran and active-duty populations. Additionally, José Gutiérrez-Maldonado contributed 28 articles, primarily addressing VRET applications for substance use and eating disorders, establishing the third most productive research trajectory in the dataset.

Patterns of author collaboration reveal a combination of individual productivity and tightly coordinated research networks. Skip Rizzo contributed 20 articles focused on clinical VR readiness and combat-related PTSD, frequently collaborating with Rothbaum and Reger to form a prominent North American military-clinical VRET research consortium; this group represents the most citation-impactful collaborative cluster identified. The collaborative structure of the field is further illustrated by Marta Ferrer-García (19 articles) and Per Carlbring (18 articles). Ferrer-García’s output is closely integrated with that of Gutiérrez-Maldonado, reflecting a productive Barcelona-based research group specializing in cue-exposure and eating disorders. Conversely, Carlbring’s work is tightly coupled with that of Per Lindner (17 articles), forming a Scandinavian research network focused on gamified and self-guided VRET for public speaking anxiety and specific phobias using consumer-grade hardware.

The remaining top-ranked authors, Stéphane Bouchard (17 articles), Greg Reger (17 articles), Sungkun Lee (17 articles), and Per Lindner (17 articles), further demonstrate the thematic diversity and geographic breadth of leading VRET scholarship. Bouchard’s research concentrated on OCD and social anxiety disorder from a Canadian institutional base, while Reger’s work extended the North American military PTSD cluster through rigorous comparative trial methodologies. The geographic distribution of the top ten authors, spanning Italy, the United States, Spain, Sweden, Canada, and South Korea, confirms that leadership in VRET research is internationally distributed across distinct thematic specializations. This global footprint suggests that the field is not concentrated within a single national infrastructure but is instead shaped by interdisciplinary and cross-cultural clinical demands.

The international co-authorship rate of 3.92%, calculated using fractional counting of institutional affiliations per article, is substantially below norms reported for clinical psychology (approximately 15%–20%) and digital health research (approximately 10%–15%). This low rate likely reflects a combination of factors: the specialist technical infrastructure required for VRET research limiting participation to well-resourced centers; the historically single-institution trial design culture of clinical psychology and psychiatry; and potential database affiliation disambiguation limitations that may undercount multi-country collaborations where author affiliations are incompletely recorded in Scopus and Web of Science metadata (Donthu et al., 2021).

The science mapping analysis of the 1,583-article dataset identified distinct thematic clusters within VRET research, encompassing clinical application domains, methodological frameworks, and participant characteristics. As shown in Figure 6, the keyword co-occurrence network revealed three interconnected clusters that reflect the multidimensional nature of VRET scholarship. Network analysis based on betweenness centrality metrics demonstrated that core VRET concepts function as the primary conceptual bridges linking these three thematic domains.

Cluster 1 (Red nodes), centered on “virtual reality exposure therapy” and “virtual reality,” constitutes the largest thematic group and covers core clinical applications and their immediate therapeutic contexts. This cluster exhibited the highest betweenness centrality values across the network, with human (0.978), virtual reality exposure therapy (0.959), and virtual reality (0.892) serving as the three essential bridges connecting different research sub-domains.

Key nodes such as humans (0.679), procedures (0.280), anxiety (0.120), and psychology (0.119) indicate a strong integration between immersive technologies and established clinical psychology frameworks. In contrast, disorder-specific keywords, including posttraumatic stress disorder, phobia, cognitive behavioral therapy, and systematic review, showed lower centrality values. This suggests they function as terminal application nodes for specific clinical niches rather than broad conceptual bridges. The inclusion of meta analysis, RCT, and outcome assessment confirms that Cluster 1 represents the empirical clinical evidence base of VRET, where methodological rigor and disorder-specific measurement are the dominant concerns.

Cluster 2 (Blue nodes), centered on “female,” “male,” and “adult,” represents a methodological and participant characterization cluster. This group reflects the demographic and study design reporting conventions prevalent in VRET clinical trial literature. Substantial betweenness centrality was observed for adult (0.312), female (0.309), and male (0.271), highlighting their roles as bridging descriptors that connect clinical findings across diverse populations and designs. Keywords such as randomized controlled trial, controlled study, treatment outcome, and follow up demonstrate that this cluster captures the methodological infrastructure of the field, specifically the sampling and reporting standards, rather than its thematic content. Furthermore, the co-occurrence of physiology and pilot study suggests an emerging methodological strand that incorporates physiological outcome measures into early-phase clinical investigations.

Cluster 3 (Green nodes), centered on “pain,” “analgesia,” and “pain management,” identifies a clinically distinct but peripherally connected research stream addressing VR-based pain intervention, particularly in pediatric and adolescent populations. This is evidenced by the co-occurrence of child and adolescent within the cluster. Although betweenness centrality values for Cluster 3 were modest, ranging from 0.011 for pain management to 0.022 for adolescent, their separation from the core anxiety and PTSD literature of Cluster 1 indicates that VR-based pain management is a recognizable but partially isolated research strand. Its positioning suggests that while this research is indexed within the same bibliometric corpus, it operates with a distinct theoretical and clinical vocabulary, resulting in reduced connectivity with core VRET concepts.

As shown in Figure 7, thematic map analysis offers a complementary perspective on the intellectual structure of the VRET field, categorized across two primary dimensions: relevance degree (centrality) and development degree (density). The current analysis delineates four distinct quadrants to categorize the developmental maturity of various research themes across the literature. The resulting pattern indicates a field where the core intervention concepts have attained high centrality, yet internal specialization remains uneven across different research streams.

Motor Themes (upper-right quadrant) were notably absent in this analysis. This finding suggests that the VRET field currently lacks research areas that simultaneously possess high internal cohesion (density) and robust external connections to other domains (centrality). The lack of motor themes is significant, as it indicates that while VRET research is well-established and expanding, no single thematic area has yet reached the dual status of being both a highly specialized research front and a central connector for the broader literature. Such a pattern is characteristic of a maturing field that has not yet consolidated around a single, dominant integrative paradigm.

Basic Themes (lower-right quadrant) contained the most central cluster of the study, with “virtual reality exposure therapy,” “human,” and “humans” identified as the core transversal concepts. Despite having relatively lower density, these themes exhibited the highest centrality values in the entire map. This confirms their role as the fundamental conceptual anchors that link multiple research streams without being confined to a single specialized niche. This positioning indicates that VRET as an intervention modality and its focus on human participants are mature, widely adopted foundations that serve as the primary connective tissue across disorder-specific, methodological, and population-focused research.

Niche Themes (upper-left quadrant) included “review,” “systematic review,” and “randomized controlled trial (topic),” representing specialized but peripherally connected methodological areas. These themes showed comparatively high internal development, reflecting a substantial and coherent body of VRET evidence synthesis and trial methodology, but limited centrality within the broader thematic network. This suggests that the scholarly sub-community focused on VRET trial methodology and evidence synthesis is well-developed and internally consistent, yet remains partially isolated from the applied clinical and technological dimensions of the field.

Emerging or Declining Themes (lower-left quadrant) covered “virtual reality,” “anxiety,” and “therapy,” all positioned with both low centrality and low density. The location of these seemingly foundational terms in this quadrant is a striking finding. It indicates that as standalone keywords, they either represent emerging conceptual directions that have not yet developed coherence, or more likely, declining indexing conventions. These broader terms appear to be superseded by the more precise descriptor “virtual reality exposure therapy,” which has successfully migrated into the Basic Themes quadrant. This terminological transition from general terms like virtual reality and anxiety toward the specific virtual reality exposure therapy reflects the conceptual maturation and increasing disciplinary precision of the VRET research community throughout the 2015–2025 period.

The analysis of trending topics illustrates the temporal progression of keyword usage from 2015 to 2025, highlighting distinct developmental phases within the VRET literature (Figure 8). The initial period (Q1 and median 2015–2016) was characterized by terms such as withdrawal syndrome, computer interface, user-computer interface, and computer simulation. This reflects the field’s early technical and addiction-centric focus, where research was primarily concerned with VR delivery infrastructure and substance-related cue exposure. Furthermore, the early prominence of feeding behavior (median 2016) confirms that eating disorder applications were among the first established clinical domains, aligning with the foundational work of Riva et al. (2019) noted in the productivity analysis.

A subsequent transitional phase (median 2017–2020) saw the consolidation of core clinical and methodological terminology. During this time, cognitive therapy (frequency: 70, median: 2017), phobic disorders (frequency: 70, median: 2019), fear (frequency: 161, median: 2020), and physiology (frequency: 129, median: 2020) became increasingly central. The rise of the term video game (frequency: 61, median: 2020) suggests the growing impact of gamification and self-guided delivery models, particularly for phobia treatments using consumer hardware. The terms human (frequency: 1,085, median: 2021) and humans (frequency: 893, median: 2021) reached their median frequency during this interval, reflecting the surge in human-participant clinical trials that began in 2019.

The most recent phase (median 2022–2025) is defined by the dominance of virtual reality exposure therapy (frequency: 1,142, median: 2022) and virtual reality (frequency: 1,140, median: 2022) as the most frequent descriptors. This confirms the terminological stabilization identified in the thematic map. Significantly, several clinical terms reached their median frequency only after 2022, including depression (frequency: 169, median: 2023), pain (frequency: 153, median: 2023), randomized controlled trial (frequency: 189, median: 2023), and therapy (frequency: 284, median: 2024). This indicates that the most active research frontiers currently involve rigorous trial methodologies and the expansion of VRET into pain and depression management. Finally, the emergence of terms like etiology (Q1: 2024), demographics (Q1: 2024), and simulator sickness questionnaire (median: 2025) suggests three nascent directions: the mechanistic study of disorder origins, demographic moderator analysis, and the standardisation of cybersickness measurement as a key safety outcome.

The qualitative analysis of the 40 studies primarily reveals seven clinical application domains where VRET demonstrates transformative therapeutic potential. These domains represent specific areas of application with varying levels of empirical maturation, ranging from well-established evidence bases supported by meta-analytic confirmation to emerging applications requiring further controlled investigation.

The analysis uncovered intricate configurations of technological integration that extend VRET capabilities well beyond simple stimulus presentation, encompassing hardware platform evolution, artificial intelligence (AI) integration, digital twin frameworks, closed-loop biofeedback systems, and telehealth delivery infrastructures that collectively define the current technological frontier of immersive therapeutic intervention.

The qualitative synthesis revealed that VRET operates through several interacting therapeutic mechanisms that are both theoretically grounded in established exposure therapy frameworks and uniquely amplified by the properties of immersive virtual environments. Understanding these mechanisms and the outcome variables used to measure them is essential for interpreting the heterogeneity of effect sizes observed across clinical application domains and for guiding the methodological standardisation that the field currently requires.

The challenges and barriers discussed in this section come from the combined evidence of the qualitative synthesis. This is supported by broader literature when specific implementation details were not covered in that primary group. Statements about clinical safety, how often cybersickness occurs, and the necessary skills for therapists are based on findings reported in several of the included studies. These points are kept separate from any speculative guesses throughout the text.

Despite the substantial and growing evidence base supporting VRET efficacy across multiple clinical domains, the translation of laboratory-validated protocols into routine clinical practice remains constrained by a constellation of challenges spanning therapeutic, technical, and regulatory dimensions. The qualitative synthesis identified these barriers as the primary factors limiting the scalability and accessibility of VRET beyond specialist research centers, and their systematic resolution constitutes one of the field’s most pressing priorities for the coming decade.

This review identified three linked research gaps and future priorities that set the agenda for the next phase of VRET study. These findings have direct meaning for doctors, researchers, and tech developers as the field moves from specialized research into everyday clinical use.

Even with the huge increase in VRET publications from 2015 to 2025, the actual evidence is still limited because most studies are small-scale trials with follow-up periods that rarely last longer than 6 months. Most of the RCTs analyzed in the qualitative review had fewer than 100 participants. This size is too small to prove how long VRET-based learning lasts, find out what keeps the treatment working over time, or see how different patients respond based on how severe their condition is (Carl et al., 2019; Emmelkamp and Meyerbröker, 2021). For clinicians, this means they cannot yet give patients long-term recovery predictions with the same confidence they have for short-term results. Future trials should make 12-month follow-up checks and larger groups of participants a standard requirement. This would allow for better analysis based on the severity of the disorder, the type of hardware used, and how the therapy was delivered.

The variety of different measurement tools used in published VRET studies is a major methodological hurdle. This inconsistency prevents the precise data grouping needed to create official clinical guidelines and makes it difficult for doctors to compare results across different trials or disorders (Wiebe et al., 2022).

The field quickly needs a agreed-upon “core outcome set” for VRET. This would list mandatory areas to measure, such as how much a patient’s fear decreases, changes in avoidant behavior, the feeling of “being there,” physical reactions, and overall quality of life. These measures should apply to all types of disorders and ways the therapy is delivered.

For researchers, using a standardised framework like this when designing trials would make it much easier to combine data in future studies. For doctors and healthcare providers, the most important step for moving VRET into regular use is the creation of specific clinical practice guidelines. These guidelines should be officially backed by professional psychology and psychiatry organizations to ensure the technology is used confidently outside of specialized research centers.

The current research on VRET is mostly focused on adults in wealthy, specialized medical centers. This leaves children, the elderly, and people in low-resource areas as underserved groups. We still do not fully understand how well these populations tolerate VRET, what cultural changes are needed, or how practical it is to implement the therapy in those settings.

Using VRET for children is a strong idea because gamified virtual worlds can keep younger patients engaged, but there are still very few large-scale trials for this age group (Lindner et al., 2019). Older adults face different challenges, such as a higher risk of cybersickness, less familiarity with the technology, and age-related sensory issues. These factors require specially designed treatment plans that the current research has not yet addressed.

For researchers, the priority should be forming partnerships with underrepresented communities, including healthcare providers in lower-income countries, pediatric specialists, and geriatric psychiatrists. This would allow for the creation of VRET programs that fit the specific context rather than just copying methods used for adults in wealthy nations. For technology developers, the main goals for global fairness should be creating affordable, standalone hardware with simple interfaces, culturally relevant content, and delivery models that can be used by healthcare workers who are not specialists.

The integration of AI is the next big step for VRET, affecting both how therapy is delivered and how it is scientifically proven. In the short term, the most exciting opportunity lies in AI-driven personalised systems. These use “reinforcement learning” to change virtual content on the fly, monitor a patient’s physical stress levels in real-time to adjust difficulty, and create personalised avatars. However, before these can be used regularly, they need to be tested in head-to-head trials against traditional, therapist-led programs to ensure they are safe and effective (Philippe et al., 2022; Bergsnev and Sánchez Laws, 2026). Looking further ahead, the “next frontier” is Digital Twin integration. This involves creating a continuous digital model of a patient that tracks their clinical, physical, and behavioral data over time to suggest exactly how their treatment should progress.

The lack of health economic evaluations is arguably the most urgent practical gap in the field. Without clear evidence of cost-effectiveness, it is difficult for insurance providers and healthcare systems to make informed decisions about reimbursement or service priorities. Currently, almost no published studies provide the specific financial data, such as incremental cost-effectiveness ratios or quality-adjusted life year (QALY) estimates, needed for VRET to compete for funding against traditional medications or standard talk therapies (Donker et al., 2019; Carl et al., 2019).

To wrap things up, achieving these priorities requires a shift toward interdisciplinary collaboration. We need a framework that brings together clinical psychologists, psychiatrists, software engineers, hardware developers, and AI researchers into shared programs with synchronized timelines. Nowadays, these groups often work in “silos.” Technological breakthroughs in VR and AI frequently happen without clinical input, while clinical trials often use whatever tech is available without optimizing it for therapy. This has led to a gap where what the technology can do and what the clinical evidence proves are often misaligned (Emmelkamp and Meyerbröker, 2021; Philippe et al., 2022).

Recent investigations highlight a shift toward more sophisticated, technology-enhanced interventions that address long-standing limitations in the field. For instance (Bergsnev and Sánchez Laws, 2026), demonstrate the transformative potential of AI in automating therapeutic adjustments, while (de Haart et al., 2026) explore how augmenting virtual environments with real-time feedback can enhance patient engagement and treatment efficacy. Complementing these technological strides (Nedungadi et al., 2025), provide critical insights into the learning mechanisms underlying successful virtual exposure, offering a clearer blueprint for optimizing clinical protocols. Furthermore, the work of (Medeiros et al., 2026) in mapping the current landscape of VRET research underscores the necessity of standardising these emerging tools to ensure they can be effectively scaled across diverse clinical populations. Together, these studies represent a necessary move toward a more personalised, data-driven framework for virtual reality-based therapy.

This analysis is subject to several limitations regarding its scope and coverage. Although the dual-database approach ensured high-quality data, the exclusion of other databases may have resulted in the absence of relevant VRET research published in more targeted clinical psychology journals, psychiatric conference proceedings, or newer digital health periodicals not yet fully indexed by these platforms. Furthermore, the decision to restrict the search to English-language publications may have limited the geographic and cultural breadth of the data, particularly in light of the significant VRET research output from non-Anglophone regions identified in the collaboration network analysis.

The qualitative synthesis utilized a citation-based selection method, which naturally favors older, established publications with more time to accumulate citations. Consequently, while the quantitative bibliometric analysis included 281 articles from 2025, none of these reached the citation threshold for qualitative review. This means that highly recent and potentially influential contributions regarding AI integration, digital twin frameworks, and telehealth delivery may be underrepresented in the clinical evidence synthesis. This methodological approach facilitates a dual focus: the qualitative synthesis examines the established intellectual structure of the domain, whereas the quantitative analysis serves as the essential instrument for capturing the latest bibliometric shifts and early-stage frontiers.

Additionally, inherent limitations in author disambiguation, common in bibliometric studies relying on surname and initial matching, may have introduced minor inaccuracies in productivity and collaboration metrics, especially for common surnames shared by multiple researchers. Finally, the rapid pace of innovation in VR hardware and software means that technical constraints or implementation hurdles identified in earlier studies within the 2015–2025 window may not accurately reflect current state-of-the-art capabilities. As the evidence base expands, continuous monitoring of the evolution in consumer and clinical VR infrastructure will be essential to maintain an accurate understanding of the field.