The Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) is an expanded version proposed by (Venkatesh et al. 2012) based on the original Unified Theory of Acceptanceand Use of Technology (UTAUT) (Venkatesh et al., 2003), specifically targeting consumer technology acceptance behavior. This model includes seven core independent variables: performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, price value, and habit. Existing empirical studies indicate that this model explains up to 70% of the variance in usage intention (Venkatesh et al., 2016). The UTAUT2 model is effective in explaining the willingness and behavior of different groups to adopt innovative technologies across various cultural environments and social contexts, and has become an important theoretical tool for predicting and analyzing the acceptance of information technology, virtual reality technology, etc., by individuals or organizations in learning contexts (Kosiba et al., 2022).

This study adopts UTAUT2 as its foundational framework. While this theory effectively predicts initial intention to adopt a technology, its explanatory mechanism for the internalization and transformation process—specifically, how intention translates into sustained and in-depth actual behavior—remains underdeveloped. To address this theoretical gap and align with the specific context of adolescents' digital sports participation, the model has been appropriately streamlined and enhanced by incorporating the theory of technological affordances as a core supplementary perspective, thereby strengthening its theoretical foundation.

First, the UTAUT2 model (Venkatesh et al., 2012) was contextualized and simplified for this study. We retained the five core antecedent variables—performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation—which constitute the initial psychological drivers of technology adoption. Considering that adolescents primarily use free or low-cost applications and are still in the formative stage of behavioral patterns, the variables of “price value” and “habit” were excluded. This refinement allows the model to focus more closely on the key driving mechanisms during the early and middle stages of participation.

Second, and central to the theoretical innovation of this model, is the introduction of the key mediating variable: virtual–real integration. The conceptualization of this construct is grounded in the theory of technological affordances. Originating from Gibson's ecological psychology (Gibson, 1979), affordances refer to the possibilities for action that an object offers to an actor, with the realization of these possibilities closely tied to the actor's perceptual capabilities. The cognitive essence lies in understanding how users perceive and apply technological features—that is, any affordance must be perceived by the actor (Norman, 1988). In the fields of organization and information systems, technological affordances have been defined as the possibilities for action or change offered by information and communication technologies to individuals or organizations (Leonardi and Vaast, 2017), as well as the process of realizing these possibilities (Du et al., 2019). The theory emphasizes that the value of technology lies not in its inherent qualities, but in the dynamic relationship through which it is perceived and utilized by actors in specific contexts to create value. Based on this, virtual-real integration is operationally defined in this study as: the process through which adolescents actively perceive and utilize the possibilities for action provided by digital technologies during digital sports participation, systematically connecting and integrating them with real-world physical activities to create coherent behavioral value. It extends beyond the typical adoption or usage focus of UTAUT2, concretely representing the critical transformative pathway from technology acceptance to behavioral deepening. In other words, the antecedent variables not only directly influence behavior but also indirectly drive meaningful participation by fostering this deep integration.

Furthermore, to accurately capture different dimensions of behavioral outcomes, this study distinguishes between two outcome variables: active sports participation focuses on the external manifestation and quantitative engagement at the behavioral level, reflected in the proactiveness, frequency, planning, and self-management behaviors exhibited by individuals when using digital sports technologies. It captures the quantity and routinization of participation. Deep sports participation emphasizes the internal state and qualitative immersion at the experiential level. Its core characteristics include high emotional engagement, deep cognitive focus during physical activity, and the internalization of value and sense of accomplishment derived from sports, drawing theoretical grounding from flow theory. It reflects the quality and sense of meaning in the participation experience.

In summary, the theoretical model constructed in this study systematically examines how the UTAUT2 antecedent variables ultimately influence adolescents' sports participation—from behavioral expression to experiential depth—through the behavioral transformation mechanism of virtual–real integration. This work not only serves as an empirical test of UTAUT2 in the context of adolescent digital sports but also represents an important extension of its theoretical boundaries. It explicitly offers a testable mediating mechanism that explains how technology adoption leads to deep behavioral change, thereby directly addressing the framework's limitations in explaining behavioral transformation processes.

Based on the aforementioned hypotheses, the theoretical model constructed in this study is shown in Figure 1. This model depicts how the five core independent variables—performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation—influence the two pathways of active sports participation and deep sports participation through the mediating variable virtual-real integration. Simultaneously, the model also incorporates gender, age, and grade as control variables to test the robustness of the relationships between the core variables after excluding the potential interference of these demographic factors. The rationale for selecting these control variables is as follows: existing research indicates that adolescents of different genders and age groups may exhibit differences in technology acceptance and sports participation behavior (Angosto et al., 2023; Zhang et al., 2023; Wang et al., 2022). Furthermore, grade level is often associated with academic pressure and discretionary time, which may also affect the depth and frequency of their sports participation.

In this research model, the independent variable is the acceptance and utilization of digital sports technology, the dependent variable is adolescent sports participation, and the mediating variable is virtual-real integration. Control variables include gender, age, and grade level. All items, excluding control variables, employ a 5-point Likert scale (1 = “Strongly Disagree” to 5 = “Strongly Agree”), with higher total scores indicating greater levels of the corresponding construct. To ensure the scientific rigor and validity of the measurement instruments, all scales underwent rigorous theoretical development and multiple rounds of revision.

Data were collected using a combination of online and offline methods. The online survey was distributed via the “Questionnaire Star”' platform, while the offline questionnaires were administered on-site during school physical education classes and youth sports club activities. All items were closed-ended questions, completed independently by the respondents. A total of 430 questionnaires were collected. After excluding 32 invalid questionnaires due to patterned responses or internal inconsistencies, 417 valid questionnaires were retained, yielding a valid response rate of 96.98%. The sample size meets the recommendation for structural equation modeling that the sample size should be at least 10 times the number of items (Wu, 2009), as this study utilized 30 items. Among the valid samples, males accounted for 49.4% (206 participants), and females accounted for 50.5% (211 participants). The age distribution was concentrated in two groups: 12–15 years old (junior high school, 51.6%) and 16–18 years old (high school, 45.4%). The sample covered students from regular classes, sports specialty classes, and extracurricular sports clubs, indicating a diverse and representative sample of the adolescent population that is suitable for subsequent empirical analysis.

This study employs Structural Equation Modeling (SEM) to test the theoretical model. SEM can simultaneously handle multiple dependent variables and permits measurement error in both independent and dependent variables, making it suitable for examining the theoretical model in this research, which encompasses multiple latent variables and their complex causal relationships. Specifically, the Bootstrap mediation effect testing method was employed (Preacher and Hayes, 2008). Through repeated sampling (set at 5,000 iterations), confidence intervals for the mediation effect were generated to examine the mediating role of virtual real integration between digital sports technology acceptance and sports participation. A mediation effect was deemed significant when the 95% confidence interval did not include zero.

The data analysis process comprised two stages: firstly, foundational statistical procedures including descriptive statistics, reliability testing, and correlation analysis were conducted using SPSS software; secondly, confirmatory factor analysis and structural equation modeling were performed using AMOS software to evaluate the measurement model's reliability and validity, alongside the structural model's path coefficients and goodness-of-fit. All analyses employed two-tailed tests with a significance level set at α= 0.05.

The results of the correlation analysis (Figure 2) revealed significant positive relationships among all core variables in this study (p < 0.01), providing preliminary support for the theoretical hypotheses. Specifically, the five antecedent variables of UTAUT2—performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation—were significantly and positively correlated with the mediating variable virtual–real integration as well as with the outcome variables active participation and deep participation. Among these, facilitating conditions showed the strongest correlation with virtual–real integration (r = 0.508), while performance expectancy exhibited the strongest correlation with active participation (r = 0.482). The mediating variable virtual–real integration also demonstrated moderately strong positive correlations with both active participation (r = 0.506) and deep participation (r = 0.468), preliminarily confirming its bridging role in the model. Furthermore, none of the correlation coefficients between variables exceeded 0.85, indicating that while the constructs are related, they maintain sufficient discriminant validity to meet the requirements for further analysis using structural equation modeling.

To systematically assess the potential influence of common method bias, this study employed two complementary statistical approaches for testing. First, a more rigorous multilevel confirmatory factor analysis was conducted for nested model comparison. The results are presented in Table 5. The baseline eight-factor model demonstrated the best model fit (χ²/df = 1.411, RMSEA = 0.031, CFI = 0.980, TLI = 0.977). As factors were successively constrained to merge in the alternative models, all fit indices showed significant and systematic deterioration. Notably, the model in which all variables were combined into a single factor yielded a very poor fit (χ²/df = 9.349, RMSEA = 0.142, CFI = 0.560, TLI = 0.524), indicating that common method variance cannot account for the majority of the covariation among the constructs. Second, the Harman's single-factor test (Podsakoff et al., 2003) was performed for cross-validation. When all measurement items were subjected to exploratory factor analysis, eight factors exhibited eigenvalues greater than 1, and the first factor explained only 35.72% of the total variance, which is below the 40% threshold. Based on the combined results of these statistical tests, it can be concluded that common method bias does not pose a serious threat to the data quality or the core findings of this study.

This study focuses on the complete influence mechanism of technology acceptance- virtual-real integration-sports participation. Testing the invariance of the integrated model allows for more direct verification of whether the measurement structure of the entire theoretical framework remains stable across different gender groups. Therefore, prior to conducting structural equation modeling, this research performed cross-gender measurement invariance tests on the model.

The results (Table 6) indicate that all levels of invariance (configurational, weak, strong, and strict) in the integrated model meet psychometric standards (Δ CFI ≤ 0.001, Δ RMSEA ≤ 0.002). This confirms that measurement parameters exhibit no significant differences between gender groups. This indicates that the core variables and their measurement relationships in this study are consistent across male and female samples.

To examine the fit between the theoretical model and the empirical data, this study utilized AMOS software to conduct a comprehensive evaluation of the overall model fit using multiple indices: absolute, incremental, and parsimonious fit.

The analysis results showed that for the absolute fit indices, the χ²/df value was 1.380 (meeting the criterion of < 3); the RMSEA was 0.030 (better than the critical value of 0.08). The incremental fit indices were excellent, with TLI = 0.973 and CFI = 0.977 (both exceeding 0.9). The parsimonious fit indices, PNFI = 0.789 and PCFI = 0.836, were both significantly higher than the benchmark of 0.5. All fit indices reached or surpassed the recommended thresholds, indicating an ideal match between the theoretical model and the observed data. The overall structure of the model is reasonable and suitable for subsequent path analysis and hypothesis testing. The results are shown in Table 7.

Based on the path analysis results obtained through the AMOS structural equation modeling, this study examined the proposed theoretical hypotheses. First, the test results concerning the control variables revealed that only age exerted a statistically significant positive effect on the mediating variable virtual-real integration(β= 0.089, p < 0.05). In contrast, gender and grade showed no significant impact on virtual-real integration, nor did any of the control variables significantly affect the dependent variables active participation and deep participation (p> 0.05). These findings indicate that the potential effects of demographic variables have been incorporated and statistically isolated in the model, with no significant interference detected in the relationships among the core constructs. Consequently, the subsequent analytical conclusions maintain a high degree of purity and reliability.

Regarding the core hypotheses, the results showed that the direct path from performance expectancy to deep sports participation was not significant (β= 0.069, p> 0.05), thus, hypothesis H1b was not supported. Additionally, the direct effects of facilitating conditions on both active and deep sports participation were not statistically significant (β= 0.033, p>0.05; β= 0.048, p>0.05). Therefore, hypotheses H4a and H4b were also not supported. (The standardized model is shown in Figure 3, and the model path results are presented in Table 8).

All other hypotheses were supported. Specifically, the factors exerting a significant positive influence on active sports participation, ranked by effect size, were: performance expectancy (β= 0.244), hedonic motivation (β= 0.170), effort expectancy (β= 0.160), social influence (β= 0.162), and the degree of virtual-real integration (β= 0.192). The factors significantly influencing deep sports participation were: social influence (β= 0.272), the degree of virtual-real integration (β= 0.206), effort expectancy (β= 0.126), and hedonic motivation (β= 0.145). Performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation all demonstrated significant positive effects on the degree of virtual-real integration, with the effect of facilitating conditions being the most prominent (β= 0.342). The model exhibited strong explanatory power for the three core endogenous variables. The squared multiple correlations (R²) for virtual-real integration, active participation, and deep participation were 0.507, 0.515, and 0.431, respectively. This indicates that the antecedent variables (performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation) collectively accounted for 50.7% of the variance in VRI, and that VRI, together with the other antecedent variables, explained 51.5% and 43.1% of the variance in AP and DP, respectively.

Based on the Bootstrap mediation effect test results (Hopwood, 2007) (Table 9), an analysis of the mediating role of the virtual-real integration in the adolescent sports participation model yielded the following conclusions: the mediation effect analysis indicates that the virtual-real integration played a significant mediating role in the relationships between performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, and the two types of sports participation behaviors. The specific results are as follows:

In the active participation in sports pathway, all five antecedent variables produced significant indirect effects (p < 0.05; 95% CIs included zero) through the degree of virtual-real integration. Furthermore, with the exception of facilitating conditions, the direct effects of the other variables on active participation were also significant. The direct effect of facilitating conditions was not significant (β = 0.033, p = 0.526), indicating that its influence is entirely mediated by the degree of virtual-real integration. That is, facilitating conditions promote active participation only by enhancing virtual-real integration, with no direct effect present.

In the path predicting deep participation, the direct effects of both performance expectancy (β = 0.069, p = 0.280) and facilitating conditions (β = 0.048, p = 0.495) were not statistically significant. The direct effect of effort expectancy approached significance (β = 0.126, p = 0.052), suggesting a potential direct promoting effect, although the statistical evidence for this path is weaker than that for other significant paths. Effort expectancy, social influence, and hedonic motivation each exhibited either significant or marginally significant direct effects, while also demonstrating significant indirect effects mediated by virtual-real integration.

Regarding the total effects, the Bootstrap results indicated that the total effects of performance expectancy and facilitating conditions on deep sports participation were not statistically significant (p = 0.134 and p = 0.075; 95% CIs included zero). In contrast, the total effects of all other antecedent variables on both active and deep participation were significant (p < 0.05), with the exception of the total effect of facilitating conditions on active participation, which was marginally significant (p = 0.053). Overall, these results indicate that the integrated model possesses good predictive power for explaining adolescents' digital sports participation behaviors. The final path coefficients of the model are shown in Figure 4.

Based on the Unified Theory of Acceptance and Use of Technology (UTAUT2), this study constructed a theoretical model to analyze adolescents' digital sports participation behavior. Based on structural equation modeling of questionnaire data from 417 adolescents, this study systematically investigated the influence of five core variables (performance expectancy, effort expectancy, social influence, facilitating conditions, and hedonic motivation) on adolescents' active and deep sports engagement, as mediated by the key construct of virtual-real integration. Findings indicate that the core constructs of the UTAUT2 model possess general explanatory power in adolescent digital sports participation contexts, yet their influence pathways and effectiveness exhibit significant differentiation.

The core contribution of this study's theoretical model lies in establishing and validating the central mediating mechanism of virtual-real integration between technology acceptance and sports participation behavior. Mediating effect tests indicate that, while facilitating conditions exert their influence entirely through this variable, the impact of performance expectations and effort expectations on deep engagement is also fully mediated by it. Hedonic motivation and community influence exert their effects partially through this variable.

This finding reveals the intrinsic logic of adolescent digital sports participation: technological factors (performance expectancy, effort expectancy and facilitating conditions) and psychosocial factors (social influence, hedonic motivation) first jointly influence the breadth and depth of adolescents' connections between virtual spaces and real-world sports. This behavioral pattern of virtual-real integration then becomes the direct driving force stimulating their active participation and deep engagement. The degree of virtual-real integration thus forms a crucial bridge connecting technological acceptance willingness with actual sports behavior, explaining the internal process by which willingness translates into action.

In terms of theoretical contributions, this study offers three key insights: First, it successfully applies the UTAUT2 model to the emerging field of youth digital sports, validating the framework's applicability and explanatory power in this context. Second, it proposes and validates the key mediating construct of virtual-real integration, offering a mechanistic explanation for how technology acceptance translates into deep sports engagement. This addresses the limitations of classical technology acceptance models in explaining the intention-to-behave conversion mechanism. Third, it reveals significant differentiation in the influence pathways of different antecedent variables on sports participation behavior, providing a more refined theoretical perspective for future research.

In terms of practical implications, this study offers specific recommendations for different stakeholders: For technology developers, focus should be placed on designing features that naturally bridge online feedback with offline exercise, such as automatically generating training plans based on exercise data or setting up virtual achievements that unlock physical rewards. For schools and physical education professionals, leveraging community influence is crucial. This can be achieved by organizing blended online-offline sports challenges and establishing exercise data-sharing communities to foster positive social comparison and mutual motivation among adolescents. For policymakers, promoting youth sports participation requires more than providing hardware. Greater emphasis should be placed on training and support to help adolescents effectively integrate digital tools with physical activities, rather than merely offering access opportunities.

This study has the following limitations: First, the cross-sectional design limits the strength of causal inference. Future research could employ longitudinal tracking or experimental designs to further validate causal relationships between variables. Second, the sample originates from a specific region, and the generalizability of conclusions requires testing across different cultural contexts. Additionally, the measurement of virtual-real integration remains open to refinement. Future work could employ multi-method validation by integrating objective behavioral data (e.g., app usage logs, motion sensor data). Finally, this study simplified the UTAUT2 model for contextualization. Future research involving broader populations or mature application scenarios may consider reintroducing variables such as price value and habit to test the model's completeness and boundary conditions.