Augmented reality (AR) is widely recognized as a highly promising digital technology for enhancing the museum visitor experience (Zhuang et al., 2021; Li et al., 2024). High-quality AR fosters immersive experiences that, in turn, enhance users’ perceptions of usefulness and ease of use. Aesthetic immersion also promotes escapist immersion, offering valuable insights for effective AR design in museums (Cheng et al., 2024).Memorable tourism experiences significantly boost place attachment through hedonic and eudaimonic wellbeing, underscoring the importance of positive psychological factors in fostering loyalty and revisit intentions (Vada et al., 2019). High-quality AR in museums significantly enhances both utilitarian and hedonic value, leading to more engaging tour experiences and improved psychological wellbeing. Content quality, system quality, and vividness are key factors driving AR adoption, highlighting the importance of well-designed AR for enriching visitors’ mental health and overall museum experience (Ma et al., 2024).

In this study, the selection of variables was primarily guided by the Information Systems Success Model (DeLone and McLean, 2003), which highlights the importance of information quality and its downstream effects on user behavior and satisfaction. Given the study’s objective to examine psychological dimensions and their impact on learning motivation and academic achievement, this model was deemed more suitable. While other well-established frameworks, such as the Technology Acceptance Model (Venkatesh and Bala, 2008) and the Unified Theory of Acceptance and Use of Technology (Venkatesh et al., 2012; Tamilmani et al., 2021), include constructs like perceived ease of use, usefulness, enjoyment, and self-efficacy, these models primarily focus on technology adoption and initial acceptance, making them less aligned with the current study’s goals. A study has revealed that user satisfaction with Instagram is influenced by two primary factors: the strength of word-of-mouth recommendations and the quality of the information shared on the platform (Danniswara et al., 2020). Zha et al. (2018) found that focused immersion plays a positive moderating role in the relationship between information quality and informational fit-to-task. Kim and Lee (2021) study revealed a significant influence of online class quality on immersion. Moreover, they found a positive relationship between online class quality and learning satisfaction. An et al. (2021) discovered that two essential characteristics of virtual travel, namely the sense of immersion and the quality of information, were identified and shown to exert a beneficial influence on tourists’ flow experience.

The information richness theory, alternatively referred to as the media richness theory (Daft and Lengel, 1986), conceptualizes communication channels as objective characteristics that dictate their capacity for information transmission. This theory encompasses four essential attributes, including the promptness of feedback, the capacity to convey multiple cues, linguistic diversity, and the likelihood of personal attention. Face-to-face communication is considered richer in information as it is bidirectional and can help alleviate comprehension challenges. However, in the context of museum studies, communication typically involves one-way persuasive communication. Chen and Chang (2018) suggested that websites have evolved significantly in terms of media richness, evolving from static textual and graphical elements to dynamic and interactive virtual experiences. This evolution aims to enhance satisfaction and confidence. Chesney et al. (2017) research revealed that the information richness offered by the virtual world interface has the potential to enhance e-commerce transactions by cultivating trust between trading partners. Jiang et al. (2022) utilized a structural equation model to explore the determinants of users’ continuance intention with museum AR technology and uncovered that information richness exerts a favorable impact on information quality.

AR has been shown to be an effective tool for presenting various museum collections, including painting guides (Ding, 2017), narrative interactions, stories behind statues (Keil et al., 2013), and learning experiences of antique relics (Khan et al., 2021). Research conducted in the field of AR applications in art museums indicates that these technologies offer visitors a more profound understanding of the intricate details and wealth of information contained in paintings, surpassing the capabilities of traditional guides, by offering more information and better information quality (tom Dieck et al., 2018). Visitors equipped with AR guides exhibit increased focus and engagement with museum artworks, ultimately leading to enhanced learning efficiency and a more immersive flow experience (Chang et al., 2014). Visitors have expressed a desire for AR devices to provide high-quality information and better learning experiences (Jiang et al., 2022).

Additionally, several researchers have examined the relationship between information quality and information richness. Aljukhadar and Senecal (2017) uncovered that the perceived information quality among participants who had recreational browsing intentions was significantly influenced when product details on an e-commerce website were presented through streaming video. Zhao et al. (2017) indicated that the rich media on websites had a positive impact on consumers’ perception of information quality. Jiang et al. (2022) undertook a study examining the elements that shape users’ intention to continue utilizing AR technology in museums, revealing that information richness positively contributes to information quality. Therefore, this study proposes the hypothesis that:

Immersion can be perceived in different ways, either as an aspect of the technology itself or as a physical or psychological reaction. It can be a bodily sensation, similar to being immersed in water, or a mental state, such as being absorbed in a good book (Witmer and Singer, 1998). Museums aim to provide immersive and memorable participatory experiences to visitors (Vesci et al., 2020). The modern museology highlights the significance of fostering visitors’ active participation and engagement, moving beyond mere displays of valuable cultural artifacts (Savenije and De Bruijn, 2017). Immersive experiences are crucial in enhancing the overall satisfaction (Scholz and Smith, 2016). Kim and Lee (2021) pinpointed the challenges of online education from the viewpoint of learners and observed a positive correlation between immersion and academic achievement. Georgiou and Kyza (2018) investigated the possible effect of student motivation on the impact of immersion in learning environments utilizing location-based AR technologies. Their study revealed a positive relationship between the level of immersion and conceptual learning gains. Huang et al. (2021) conducted a study to investigate how learners’ motivation, engagement, performance, and spatial reasoning evolve over time, taking into account varying degrees of immersion. The findings indicated that higher levels of immersion were associated with increased levels of motivation and engagement among learners. Cheng and Tsai (2020) confirmed that students’ perceived immersion in a virtual learning environment is associated with their learning perceptions and serves as a predictor of positive affective outcomes.

Burdea and Coiffet (2003) define imagination as a characteristic elicited by the content of virtual environment applications, wherein the user’s mind possesses the ability to perceive non-existent entities or concepts. Madini and Alshaikhi (2017) discovered that the variable of imagination held significant importance for high-immersion VR systems that utilized head-mounted displays. Huang et al. (2010) found that virtual reality learning environments have the capability to simulate the real world, thereby stimulating the learner’s imagination and enhancing their capacity for conceptualization. Creative imagination grants learners the ability to mentally envision novel ideas and concepts that are not immediately discernible through their senses (Singer, 2000), thereby facilitating users in accomplishing their learning goals (Huang et al., 2010). Burdea and Coiffet (2003) argue that the assessment of user attitudes and receptiveness toward learning system acceptance should incorporate the factor of imagination. Huang et al. (2010) study revealed that virtual learning environments that stimulate imagination serve as effective tools for enhancing learners’ problem-solving abilities. Huang et al. (2016) performed a research investigating learners’ acceptance of desktop and projection-based display systems in the context of medical education. The research revealed that imagination had a positive impact on learners’ perception of the usefulness and ease of use of virtual learning environments, ultimately influencing their behavioral intention to adopt such systems. Therefore, this study proposes the hypothesis that:

Several researchers have highlighted the significance of academic achievement as an essential construct, emphasizing its role as an educational outcome (Choi and Kim, 2013). Academic achievement is commonly defined as the extent to which students accomplish educational objectives. It typically encompasses attaining specific outcomes in online assignments and assessments, often quantified as a grade point average (GPA) or a numerical ranking (Chowa et al., 2015). Two key approaches prevail in the study of academic achievement. Firstly, the quantitative approach relies on the assessment of students’ grade point average (GPA). Alternatively, an alternative method employs a knowledge acquisition and achievement framework to uncover abstract influencing factors (Choi and Kim, 2013). According to some studies, it has been suggested that students, especially those in lower grades, may tend to overestimate their own performance in a social expectation report, which can potentially impact the validity of the findings. In future research, it is recommended to utilize actual online results of students instead of relying solely on the online scores reported by the students themselves. This approach would help eliminate potential social biases (Broadbent and Poon, 2015).

Prior studies have investigated the relationship between learning motivation—comprising of both internal and external motivations—and GPA, which is used as an indicator of academic performance (Choi and Kim, 2013). A cross-lagged regression model was employed to analyze the interrelationships between academic achievement and motivation among high school students. The findings revealed that autonomic motivation, which measures the level of relative autonomy over the past year, exhibited a positive correlation with academic achievement even after accounting for baseline achievement (Taylor et al., 2014). This discovery underscores the crucial role of learning motivation as a key predictor of academic achievement. The research indicates that students who are deeply engaged in the joy and challenges presented by learning motivation are more likely to attain greater academic achievements (Choi and Kim, 2013). Su and Cheng (2019) investigated the impact of a virtual chemistry laboratory, utilizing the sustainable innovation experiential learning model, on academic achievement. The research outcomes demonstrated a favorable correlation between learning motivation and academic achievement, as revealed through survey analysis. Therefore, this study proposes the hypothesis that:

The term “wearable comfort” is defined as the overall subjective assessment of the physical sensation experienced by consumers when wearing AR glasses (Knight and Baber, 2005). Kuru and Erbuğ (2013) introduced the construct of “wearability,” which refers to the specific characteristics of a device, including its ease of carrying and how well it fits with the human anatomy. Kress et al. (2020) argues that two fundamental factors that define the major challenges in mixed reality hardware are comfort and immersion. In their qualitative study of Google Glass, tom Dieck et al. (2016) observed that users frequently commented on the weight of the devices. In the context of AR and VR Smart Glasses, Rauschnabel (2018) posits that the comfort of wearable devices is influenced by a range of their physical attributes, including factors such as weight, bulkiness, operating temperature, and physical pressure experienced while wearing the glasses. Herz and Rauschnabel (2019) formulated and evaluated a comprehensive framework to analyze consumer reactions toward wearable VR glasses. The study revealed that wearable comfort positively influences the attitude toward using VR glasses. Consistent with previous research, we suggest that utilizing AR glasses that provide a high level of wearable comfort results in a positive user experience. Therefore, this study proposes the hypothesis that:

The Wuhan Natural History Museum is devoted to amassing a diverse array of natural specimens, encompassing animals, plants, artifacts from the Paleolithic era, and fossils pertaining to paleoanthropology. It serves the purpose of educating the public, hosting exhibitions, and disseminating scientific knowledge. To enhance the visitors’ overall viewing experience, the museum has embraced the use of AR as a new medium. Specialized AR glasses, such as the Rokid Air, are provided to tourists to enrich their viewing experience, as shown in Figure 1, allowing them to access AR animations. The Rokid Air AR smart glasses were chosen for their lightweight design (83 g), affordability, and practical features, including a 1080p OLED display and 43° field of view. Compared to higher-cost enterprise-focused models like Magic Leap 2, Rokid Air offers an accessible and suitable option for museum applications. During their museum visits, tourists wear these AR glasses, and the AR content is activated by positional markers positioned in front of each exhibit, which are identified by corresponding signposts. The AR content included 3D animations, visual overlays, and auditory narratives, such as prehistoric creatures in motion or reconstructions of Paleolithic artifacts. Through the AR glasses, visitors gain access to visual and auditory information, seamlessly integrating traditional exhibits with cutting-edge technology, as depicted in Figure 2.

Face-to-face surveys were conducted with visitors who interacted with the AR-integrated exhibits at the Wuhan Natural History Museum. Of the 297 visitors who utilized the AR technology and completed the survey, 31 responses were excluded due to incomplete data. The minimum sample size necessary for this research was established using G*Power software to guarantee the statistical significance of the findings (Faul et al., 2009). With a medium effect size of 0.15, a desired statistical power of 0.80, and an alpha level set at 0.05, the sample size calculation considered a model incorporating a maximum of 7 predictors. Based on the calculations, a minimum of 55 respondents was necessary for the experiment. The sample size of 266 participating tourists was deemed sufficient for data analysis. Descriptive statistics of the participants can be found in Table 1.

The survey consists of two sections. The initial section assesses respondents’ level of familiarity with AR, while the subsequent section comprises inquiries relating to seven distinct measurement constructs: information quality, information richness, immersion, imagination, learning motivation, academic achievement, and wearable comfort. All measurement constructs in this study were evaluated using a multi-item 7-point Likert scale, spanning from 1 (strongly disagree) to 7 (strongly agree). The measurement constructs employed in this study were tailored from previous research, incorporating minor adjustments in wording to achieve greater congruence with the precise emphasis on AR as the subject of this research.

Participants were approached on-site at the Wuhan Natural History Museum as they expressed interest in experiencing the AR exhibits, and participation was entirely voluntary. On average, visitors wore the AR glasses for about 20–30 min; only a few reported mild discomforts, which dissipated quickly upon removing the glasses. Immediately after using the AR glasses, participants were guided to a quiet seating area within the museum to complete an electronic questionnaire, which typically took around 5–10 min. This immediate post-experience survey helped ensure that responses accurately reflected visitors’ recent interactions with the AR technology. Data collection spanned 4 weeks, during which we maintained a consistent procedure for recruiting volunteers and administering the survey, thus ensuring methodological rigor throughout the study.

The entirety of the measurement items utilized in this study were derived from previous research. Specifically, three items measuring information quality were adapted from studies conducted by Roca et al. (2006), while three items assessing information richness were adapted from studies conducted by Oh et al. (2009) and Otondo et al. (2008). Additionally, three items of immersion and three items of imagination were adapted from Barrett et al. (2021) and Bueno et al. (2020). Four items pertaining to learning motivation were adapted from Pintrich and Schrauben (1992), while four items related to academic achievement were adapted from Peterson et al. (2010). Three items of wearable comfort were adapted from studies conducted by Herz and Rauschnabel (2019) and Sukwadi et al. (2022). Ahead of administering the survey to the participants, a pilot test was executed with university students in China. Additionally, following the revisions, five researchers scrutinized the measurement items. A detailed compilation of these items is presented in Table 2.

The model underwent testing with the utilization of SmartPLS 4. According to Hair et al. (2011), Partial Least Squares Structural Equation Modeling (PLS-SEM) is a suitable method for predicting complex models that involve numerous structures, indicator variables, and structural paths, without requiring assumptions about the distribution of the data. PLS-SEM is advantageous in that it does not rely on normality assumptions and can effectively handle small sample sizes (Hair et al., 2019). Moreover, PLS-SEM is compatible with resampling methods, which are considered more effective than traditional tests like the Sobel test, particularly for analyzing indirect effects. PLS-SEM operates through two primary stages: the measurement model and the structural model. The measurement model evaluates the relationships between latent variables and their observed indicators, ensuring reliability and validity. The structural model assesses the relationships between latent variables, testing the proposed hypotheses. Additionally, PLS-SEM supports advanced features like bootstrapping and permutation tests to analyze indirect effects and enhance statistical rigor. Given its ability to handle complex models with numerous constructs and paths, PLS-SEM is widely adopted in various fields, including social sciences, marketing, and information systems. Its flexibility and focus on prediction make it a robust tool for addressing intricate research questions and providing actionable insights.

Since all the data originated from a single source, there is a potential concern regarding common method variance (CMV). In PLS-SEM, two commonly used approaches are employed to assess CMV. The initial methodology involves a thorough collinearity evaluation employing variance inflation factors (VIFs) (Kock, 2015). Following the recommendations of Kock (2015) and Kock and Lynn (2012), all variables should be regressed against a common variable. If the resulting VIF value is 3.3 or below, it signifies the absence of bias in single-source data. The satisfactory VIF values for all constructs indicate that the presence of CMV was not a significant concern in this study.

The second approach for assessing CMV employs the correlation matrix procedure. Under this approach, if the correlations among constructs remain under 0.9, it signals the absence of CMV. Our results indicate that the correlations between constructs in both groups are indeed below 0.9, strengthening the argument that CMV did not pose a significant issue in this study.

To evaluate the reliability and validity of the measurement model, two criteria were utilized: reliability and validity. To assess reliability, we calculated Cronbach’s alpha and composite reliability (CR) as measures of internal consistency, along with evaluating the outer loadings for each individual item. Table 2 provides descriptive statistics for the indicator items and factor loadings for each item. Notably, all indicator items exhibited outer loadings surpassing the recommended threshold of 0.70 (Fornell and Larcker, 1981), thus signifying satisfactory reliability.

Table 3 outlines the Cronbach’s alpha and composite reliability (CR) values for each construct. The analysis revealed that all constructs had Cronbach’s alpha values exceeding the recommended threshold of 0.70, as suggested by Hair et al. (2012). Additionally, the CR values for all constructs surpassed the established cutoff of 0.70 (Hair et al., 2012).

To assess convergent validity, the study adhered to two key criteria: evaluating the outer loadings of individual items and calculating the average variance extracted (AVE) for each construct (Fornell and Larcker, 1981). The analysis revealed that all items had factor loadings exceeding the recommended threshold of 0.70. Furthermore, the AVE values for all constructs surpassed the cut-off threshold of 0.5 (Fornell and Larcker, 1981). The results from Tables 2, 3 demonstrate satisfactory fit indices for convergent validity, indicating an adequate level of validity.

To assess discriminant validity, three distinct criteria were utilized: (1) an analysis of inter-item cross loadings, (2) the Fornell-Larcker criterion, and (3) the Heterotrait-Monotrait ratio (HTMT) criterion. These criteria were employed to ensure that the constructs in the study exhibited adequate discriminability. As indicated in Table 4, the results reveal that the indicator items measuring each construct exhibit significant and robust correlations among themselves, while also exhibiting higher loadings on their respective constructs. This serves as evidence that supports the discriminant validity of the measurement model.

Table 5 displays the Fornell-Larcker correlation matrix, which was employed to evaluate discriminant validity. The diagonal elements represent the square roots of the AVE, while the off-diagonal elements represent the estimated correlations between the respective pairs of constructs (as indicated by the rows and columns). Discriminant validity is deemed satisfactory when the square root of the AVE for each construct is consistently higher than the corresponding inter-construct correlation (Fornell and Larcker, 1981). The results in Table 5 confirm that all square roots of the AVE values exceed their respective squared correlations, thereby validating the satisfactory discriminant validity of the measurement model.

The HTMT criterion serves as a reliable metric for assessing discriminant validity by estimating the correlations between distinct constructs. As shown in Table 6, all HTMT values are well below the recommended threshold of 0.90, as advocated by Fassott et al. (2016), thereby indicating robust discriminant validity.

The thorough evaluation presented above offers compelling evidence that underscores the reliability and validity of the proposed model, encompassing both convergent and discriminant aspects of validity.

Figure 3 visualizes the proposed structural model, depicting the estimated regression path coefficients (β) and the outer loadings of the indicator items, accompanied by their respective significance levels. The figure also incorporates the moderating effect within the research model. As expected, the information richness has a significant effect on information quality (β = 0.557, p < 0.001) and on academic achievement (β = 0.239, p < 0.001). Information quality has a significant influence on immersion (β = 0.527, p < 0.001), on imagination (β = 0.489, p < 0.01), and on academic achievement (β = 0.327, p < 0.01). Immersion (β = 0.130, p = 0.003) and Imagination (β = 0.666, p < 0.001) has a significant influence on learning motivation. Learning motivation has a significant influence on academic achievement (β = 0.400, p < 0.001). Wearable comfort has positive moderating effects on the relationship between information quality and immersion (β = 0.147, p = 0.002), information quality and imagination (β = 0.141, p = 0.014).