Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice BMC Psychol . 2026 Mar 3;14:482. doi: 10.1186/s40359-026-04147-8 Search in PMC Search in PubMed View in NLM Catalog Add to search The dark side of AI transparency: investigating AI-induced anxiety, technostress, and consumer satisfaction through a hybrid SEM-ANN approach Weichen Jia Weichen Jia 1 School of Media and Law, NingboTech University, Ningbo, China Find articles by Weichen Jia 1 , Ahmed Muneeb Mehta Ahmed Muneeb Mehta 2 Hailey College of Banking and Finance, University of the Punjab, Lahore, Pakistan 3 Wittenborg University of applied sciences, Apeldoorn, Netherlands Find articles by Ahmed Muneeb Mehta 2, 3 , Chao Han Chao Han 1 School of Media and Law, NingboTech University, Ningbo, China Find articles by Chao Han 1, ✉ , Muhammad Asif Muhammad Asif 4 UE Business School, University of Education, Lahore, Pakistan Find articles by Muhammad Asif 4 , Muhammad Farrukh Shahzad Muhammad Farrukh Shahzad 5 College of Economics and Management, Beijing University of Technology, Beijing, China Find articles by Muhammad Farrukh Shahzad 5 Author information Article notes Copyright and License information 1 School of Media and Law, NingboTech University, Ningbo, China 2 Hailey College of Banking and Finance, University of the Punjab, Lahore, Pakistan 3 Wittenborg University of applied sciences, Apeldoorn, Netherlands 4 UE Business School, University of Education, Lahore, Pakistan 5 College of Economics and Management, Beijing University of Technology, Beijing, China ✉ Corresponding author. Received 2025 Oct 14; Accepted 2026 Feb 5; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13063937 PMID: 41776686 Abstract As AI-driven technologies transform customer interactions, understanding their impact on consumer satisfaction is crucial. This study investigates the relationship between artificial intelligence (AI) transparency and consumer satisfaction, incorporating AI-induced anxiety and technostress as mediators and consumer digital literacy as a moderator. A hybrid methodology combining structural equation modelling and artificial neural network (SEM-ANN) was employed to analyse data from 341 e-commerce consumers from China. The SEM results reveal that AI transparency positively influences consumer satisfaction, while AI-induced anxiety and technostress mediate this relationship negatively. However, the moderating role of consumer digital literacy is not significant. The ANN findings identify consumer digital literacy as the most influential factor, highlighting its critical role in shaping consumer responses to AI-powered interactions. These results provide novel insights into the psychological impact of AI transparency on consumers and offer theoretical and practical implications for enhancing AI-driven customer service strategies. Supplementary Information The online version contains supplementary material available at 10.1186/s40359-026-04147-8. Keywords: AI transparency, Consumer satisfaction, AI-induced anxiety, Technostress, Artificial neural network Introduction AI is rapidly becoming essential in modern consumer applications, transforming industries such as banking [ 65 ], healthcare [ 10 ], e-commerce [ 57 ], and customer service [ 62 ]. These AI systems analyse vast amounts of data to predict customer preferences, automate responses, and enhance services to boost efficiency, personalisation, and decision-making [ 34 ]. As AI technology evolves, transparency in AI operations has emerged as a vital factor influencing customer acceptance and trust [ 93 ]. AI transparency refers to how well users receive clear, understandable, and readily available information about AI systems’ processes, decision-making, and data management [ 26 ]. Upholding transparency is crucial for fostering trust, confronting ethical issues, and promoting responsible AI usage [ 71 ]. The adoption of AI across various domains is accelerating. In banking, AI chatbots and robo-advisors provide services like fraud detection and financial consultations [ 36 ]. In healthcare, AI systems assist with patient management, medical diagnostics, and tailored treatment plans [ 9 ]. Similarly, e-commerce leverages AI-driven recommendation systems to suggest products based on customer preferences and behaviours [ 91 ]. While these advancements offer unprecedented efficiency and convenience, they also introduce new challenges, particularly concerning customer perceptions and mental well-being [ 60 ]. AI transparency presents a paradoxical effect on consumers [ 87 ]. On the other hand, greater transparency fosters trust and confidence; individuals are more likely to engage with AI-driven services when they understand the decision-making processes [ 20 ]. Excessive transparency may overwhelm users with the complexity of AI decision-making [ 51 ], potentially leading to anxiety and technostress. This discomfort arises when individuals feel uncertain or uneasy about the impact of AI on their experiences, especially if they lack the technical knowledge to fully grasp its operations [ 37 ]. Similarly, customer satisfaction and the adoption of AI-based services may be hindered by technostress, which is the stress caused by technological complexity and information overload [ 46 ]. Transparency in AI is frequently seen as a crucial component in fostering consumer trust [ 16 ]. However, less is known about its unforeseen psychological impacts, such as anxiety and technostress brought on by AI. Although users can gain insight into AI decision-making through transparency, more information might overwhelm customers, resulting in cognitive overload and decreased satisfaction [ 39 ]. Current research mostly concentrates on the advantages of AI transparency [ 1 , 38 , 64 ], ignoring the possibility that it could cause tension and worry, which could have a detrimental effect on users’ experiences with AI-driven services. Therefore, research into the dual effects of AI transparency on customer satisfaction and well-being is imperative. The relationship between AI transparency, anxiety caused by AI, technostress, and customer satisfaction remains largely unexplored, especially in sectors such as e-commerce. Furthermore, most prior studies have used linear statistical models [ 1 , 16 , 20 ], which might not adequately represent the nuanced interactions among these variables. Grounded in the stimulus-organism-response (S-O-R) model, this study conceptualises AI transparency (stimulus) as influencing consumer anxiety and technostress (organism), ultimately affecting consumer satisfaction (response). To address these gaps, this study employs a hybrid SEM-ANN approach to identify both linear and nonlinear impacts, thereby providing a deeper understanding of how consumer psychology and satisfaction are influenced by AI transparency. Thus, to address these gaps, the study has addressed the following research questions. What is the relationship between AI transparency and consumer satisfaction? What is the mediating role of AI-induced anxiety and technostress in the relationship between AI transparency and consumer satisfaction? What is the moderating effect of consumer digital literacy on the relationship between AI transparency and consumer satisfaction? What are the key factors influencing consumer satisfaction using ANN? This study addresses both the advantages and unexpected drawbacks of AI transparency, offering important insights into the psychological effects on consumer behaviour. The study offers a comprehensive understanding of how AI transparency impacts consumer satisfaction by examining AI-induced anxiety and technostress as mediators, and consumer digital literacy as a moderator. By capturing both linear and nonlinear interactions, a hybrid SEM-ANN technique improves methodological rigour and provides more in-depth, data-driven insights. The results will ultimately guide AI developers, legislators, and service providers in creating user-friendly and morally sound AI systems by assisting companies in optimizing AI transparency methods to maximize consumer trust while minimizing psychological distress. The remainder of the study is structured as follows. In addition to summarising previous research, the next section describes the theoretical framework that serves as the basis for developing hypotheses. Next, the selected methodology is presented. The results of the study are shown and explained in the following section. The study’s conclusions, limitations, and potential future research directions are then covered. Theoretical framework and hypotheses development Theoretical framework Stimuli-organism-response framework According to Xu et al., [ 89 ], the S-O-R framework, which Mehrabian and Russell [ 58 ] established, consists of a series of events that begin with exposure to environmental cues (stimuli) that alter the user’s internal state (organism) and result in behavioural reactions (response). As a reliable framework to examine how consumers behave when dealing with contemporary technologies, the model has been successfully used in information systems research [ 75 ]. According to Suh and Prophet [ 81 ], the S-O-R framework is ideally suited for researching immersive technologies since it illustrates how environmental technology stimuli may affect consumers’ decision-making processes. For example, several studies applied S-O-R to investigate online reviews influencing purchase intention [ 95 ], users’ intention to use ChatGPT for learning [ 23 ], investigating binge-watching and its effect on paid subscriptions [ 80 ], and consumer avoidance toward message stream advertising on mobile social media [ 53 ]. The S-O-R framework is especially useful for comprehending how consumer behaviour and psychology are impacted by AI transparency. The stimulus (S) in this study is AI transparency, which elicits emotional and cognitive reactions, such as AI-induced anxiety and technostress (O), which in turn affect customer satisfaction (R). To strengthen the psychological grounding of this model, this study draws on cognitive appraisal theory [ 50 ], suggesting that transparency can trigger threat appraisals when users perceive information as complex or uncontrollable. Similarly, information overload theory [ 24 ] explains how excessive disclosures can overwhelm cognitive capacity, contributing to technostress. Additionally, psychological constructs such as perceived autonomy and control help clarify why users may experience anxiety when AI systems appear overly dominant or opaque. According to earlier studies, greater AI transparency can boost trust [ 20 , 93 ], but more transparency can cause psychological suffering and cognitive overload [ 39 ]. S-O-R is adopted in this study because it provides a holistic framework capable of integrating both cognitive and affective mediators, such as anxiety and technostress, unlike more narrowly focused models like the technology acceptance model (TAM) or stress-strain-outcome (SSO). Its flexibility allows for the incorporation of multiple organismic states, making it well-suited for analysing the complex psychological mechanisms triggered by AI transparency. This study presents a systematic method for analyzing the advantages and disadvantages of AI transparency using the S-O-R model. This enables a deeper understanding of how consumer digital literacy impacts these interactions. Figure 1 illustrates the proposed research model and hypotheses. Fig. 1. Open in a new tab Proposed research model Psychological foundations of AI-Induced anxiety and technostress To reinforce the psychological grounding of the S-O-R model, this study integrates key theories explaining how consumers interpret and emotionally react to AI interactions. Guided by cognitive appraisal theory [ 50 ], AI transparency is viewed as a stimulus that consumers evaluate as either reducing uncertainty or signaling a potential threat [ 30 ]. When transparency cues highlight complexity, privacy risks, or unfamiliar algorithmic logic, consumers may form threat appraisals that lead to AI-induced anxiety and technostress [ 41 ]. Research on technology-related anxiety and locus of control further explains individual differences in these organismic states. Consumers with a weaker technological locus of control may feel less capable of navigating AI systems, making them more vulnerable to stress when exposed to algorithmic explanations [ 30 ]. Although consumer digital literacy did not significantly moderate the model empirically, its inclusion remains theoretically justified because higher literacy typically strengthens perceived control. Information overload theory [ 24 ] provides an additional mechanism. Detailed transparency information can exceed consumers’ cognitive capacity, create overload, and contribute to technostress [ 43 ]. Similarly, insights from perceived autonomy and algorithmic control suggest that transparent disclosure of automated decision-making may reduce perceived autonomy, triggering discomfort and anxiety [ 55 ]. Together, these psychological mechanisms enrich the organism component of the S-O-R framework. AI transparency (Stimulus) shapes internal cognitive and affective processes (Organism), which subsequently influence consumer satisfaction (Response). This integration provides a more robust psychological justification for the mediating roles of anxiety and technostress in AI-driven service contexts. Hypotheses development AI transparency and consumer satisfaction AI transparency plays a crucial role in shaping consumer trust and satisfaction by providing clarity on how AI-driven decisions are made [ 66 ]. Transparency in AI systems enhances perceived fairness, reliability, and control, leading to greater consumer confidence in AI-enabled services [ 63 ]. However, excessive transparency can have unintended consequences, such as information overload and cognitive fatigue, which may diminish user satisfaction [ 39 ]. Study suggests that while consumers appreciate AI explanations, more technical detail can induce stress and confusion, ultimately reducing their overall satisfaction [ 61 ]. Additionally, the effectiveness of AI transparency in enhancing consumer satisfaction depends on individual differences, such as digital literacy, which influences how users interpret and respond to AI-generated information [ 13 ]. Thus, understanding the dual impact of AI transparency and its ability to build trust while potentially causing anxiety and technostress is essential for designing AI systems that optimise consumer satisfaction. Therefore, we propose the following: H1 AI transparency positively influences consumer satisfaction. AI-induced anxiety as a mediator AI-induced anxiety refers to an effective response that arises when individuals feel uncertain, threatened, or cognitively challenged by algorithmic technologies [ 79 , 84 ]. Unlike general situational anxiety, which is transient and triggered by environmental conditions, AI-induced anxiety is technology-specific and emerges when users perceive a lack of understanding, lack of control, or potential risk in AI-driven decisions. Drawing on cognitive appraisal theory [ 50 ], highly technical or extensive transparency information may be appraised as a threat rather than a clarifying cue. When AI explanations appear complex, intrusive, or error-prone, users may interpret them as signals of potential harm, thereby activating anxious responses. Excessive algorithmic information can overwhelm consumers’ cognitive resources, leading them to worry about privacy violations, incorrect decisions, or hidden algorithmic motives [ 76 , 90 ]. This aligns with research showing that AI transparency can unintentionally increase anxiety by heightening perceived uncertainty and risk [ 67 , 79 ]. Elevated AI-induced anxiety reduces trust, weakens confidence in AI recommendations, and lowers overall satisfaction [ 15 ]. Therefore, within the S-O-R model, AI transparency (stimulus) shapes internal affective states (organism), which subsequently influence consumer satisfaction (response). Thus: H2 AI-induced anxiety mediates the relationship between AI transparency and consumer satisfaction. Technostress as a mediator Technostress is a form of technology-related strain characterized by feelings of cognitive overload, psychological fatigue, and resistance resulting from complex or demanding digital interactions [ 2 , 21 ]. Unlike generic perceived stress, technostress specifically reflects the psychological pressure caused by technology’s complexity, intrusiveness, or constant demand for cognitive effort [ 74 ]. Transparency in AI is intended to enhance clarity; however, when algorithmic disclosures provide excessive technical detail or require high cognitive processing, users may experience cognitive exhaustion and information overload, two key components of technostress [ 24 , 48 ]. From a psychological perspective, transparency that is overly detailed can exceed users’ capacity to process information, triggering frustration, stress, and diminished autonomy. Such technostress impairs perceptions of usability, reduces trust, and ultimately lowers satisfaction with AI-driven services [ 56 , 69 ]. Prior research also shows that individuals with lower digital literacy are more vulnerable to these stress reactions [ 40 ]. Thus, technostress represents a key organismic mechanism through which AI transparency affects consumer outcomes in the S-O-R framework. H3 Technostress mediates the relationship between AI transparency and consumer satisfaction. Consumer digital literacy as a moderator Consumer digital literacy encompasses an individual’s capability to understand, assess, and utilize digital technologies, including AI-driven services [ 13 ]. It plays a vital role in shaping how consumers interpret and respond to AI transparency, especially in alleviating AI-related anxiety and technostress. Those with higher digital literacy are better equipped to understand AI explanations, which decreases uncertainty and boosts trust, ultimately leading to increased satisfaction [ 28 ]. Conversely, individuals with low digital literacy may experience heightened cognitive overload, making AI transparency seem daunting and contributing to anxiety and technostress [ 88 ]. Theoretically, the organism-response link is strengthened by appraisal and stress-coping perspectives, which suggest that negative affective states (e.g., anxiety, stress) directly shape evaluative outcomes such as satisfaction [ 75 ]. Thus, digital literacy operates as a boundary condition that determines whether transparency-induced reactions escalate into reduced satisfaction or are buffered through greater comprehension and perceived control. Acting as a moderating variable, consumer digital literacy influences both the magnitude and direction of the relationship between AI transparency and consumer satisfaction, underscoring the need to tailor AI explanations to different levels of user expertise [ 13 ]. In addition, digital literacy moderates the organismic pathway itself by altering the intensity with which anxiety and technostress translate into satisfaction outcomes, thus providing a more nuanced understanding of when transparency leads to positive or negative evaluations [ 88 ]. This study aims to provide insights into how businesses can refine their AI transparency approaches to enhance consumer satisfaction while minimizing psychological strain. Therefore, we assume the following: H4 Consumer digital literacy moderates the relationship between AI transparency and consumer satisfaction such that the positive effect of AI transparency on consumer satisfaction is stronger for individuals with higher digital literacy. Methodology To gather data for this study aimed at testing the proposed hypotheses, a structured questionnaire was employed. The focus was on Chinese e-commerce users who have interacted with AI-driven services, such as chatbots, automated decision-making systems, and personalised recommendations. Participants were selected using purposive sampling based on their experience with AI-powered e-commerce platforms, ensuring both relevance and data quality. This approach is effective for research requiring respondents with specific characteristics to achieve the study’s objectives [ 17 , 47 ]. A screening question was included in the survey to ensure respondents were familiar with e-commerce services. To reach potential participants, a link to a Google Forms survey was shared across various consumer forums and online platforms. The minimum required sample size, determined through statistical power analysis using the G*Power tool [ 25 , 82 ], was calculated to be 184 individuals based on four predictors, an effect size of 0.15, and a statistical power of 0.95. Ultimately, 341 responses were collected to enhance the robustness and reliability of the statistical analysis. Table 1 illustrates the demographic profile of the participants. Table 1. Participants’ demographic profile Demographic variables Category Frequency ( n ) Percentage (%) Gender Male 193 56.60 Female 148 43.40 Age group 18–24 115 33.72 25–34 138 40.47 35–44 58 17.01 45 and above 30 8.80 Educational level Intermediate 52 15.25 Graduate 113 33.14 Master 109 31.96 PhD 67 19.65 E-commerce experience (years) Less than 1 49 14.38 1–3 116 34.01 4–6 91 26.68 More than 6 85 24.93 Frequency of online purchase Rarely 109 31.97 Occasionally 117 34.31 Frequently 89 26.10 Very frequently 26 7.62 Open in a new tab There are two main components to the research instrument. Questions about the demographics of the e-commerce users make up the first section. The items measuring the five constructs in the suggested study model make up the second section (Appendix 1). A 5-point Likert scale, with 1 denoting “strongly disagree” and 5 denoting “strongly agree,” is used to score the items [ 12 , 13 ]. The five items used to measure consumer satisfaction were taken from Singh and Singh [ 78 ]. Wang and Wang [ 85 ], established five items for measuring AI-induced anxiety. A set of four items created by Fernández-Fernández et al., [ 27 ] was used to assess technostress. To measure AI transparency, Wanner et al., [ 86 ] provided three items. Asif and Sarwar [ 14 ], provided four questions that were modified to assess consumer digital literacy. A hybrid method, SEM-ANN, was employed to analyse the data. This combination has proven effective in several previous studies [ 4 , 82 ]. The primary reason for incorporating ANN alongside SEM is that SEM alone cannot adequately capture complex non-linear interactions [ 11 , 72 ]. ANN is particularly beneficial in predictive research as it analyzes data patterns to enhance prediction accuracy [ 49 ]. Furthermore, ANN provides deeper insights into the key elements that influence consumer behaviour by ranking the significance of factors impacting consumer satisfaction within AI-driven e-commerce environments [ 5 , 59 ]. In addition, this hybrid approach enables a clearer distinction between causal explanation (via SEM) and predictive importance (via ANN), ensuring both theoretical validation and predictive robustness. ANN helps uncover potential non-linear or asymmetric effects, such as threshold-based reactions to digital literacy or AI transparency, that SEM may not detect [ 94 ]. This integration, therefore, allows for a more complete understanding of consumer psychological responses to AI-driven services. In this study, SEM was conducted using SmartPLS-4, while ANN analysis was performed using IBM SPSS. Results Common method biases Data quality is seriously threatened by common method biases (CMB), which result from a single source of data collection. To lessen the possibility of CMB, procedures and statistical safeguards were implemented [ 68 ]. The intended goal of this study was stated on the questionnaire for procedural remedies. As a statistical correction, Harmon’s single-factor test was applied to every research variable [ 42 ]. The results showed that the maximum variance (37%) attributable to a single component was less than the recommended threshold value [ 68 ]. Kock [ 44 ] and [ 45 ], predicted that all items’ variance inflation factor (VIF) values in the PLS-SEM experiments would be ≤ 3.3, which would mean that the framework is free of CMB. In the current study, all the study items achieved the minimum VIF threshold values (see Table 2 ). Table 2. Reliability and convergent validity results Constructs Items FL VIF CA CR AVE Consumer satisfaction CSA1 0.808 2.413 0.896 0.923 0.706 CSA2 0.831 2.313 CSA3 0.848 2.655 CSA4 0.862 2.556 CSA5 0.852 2.447 AI transparency AIT1 0.842 1.977 0.853 0.910 0.771 AIT2 0.903 2.515 AIT3 0.889 2.052 AI-induced anxiety AIA1 0.761 2.219 0.944 0.957 0.818 AIA2 0.845 2.853 AIA3 0.979 2.327 AIA4 0.957 2.171 AIA5 0.959 1.532 Technostress TST1 0.806 2.019 0.907 0.936 0.784 TST2 0.863 2.517 TST3 0.936 1.191 TST4 0.932 1.901 Consumer digital literacy CDL1 0.819 2.010 0.902 0.935 0.764 CDL2 0.872 2.520 CDL3 0.925 1.191 CDL4 0.921 2.901 Open in a new tab Measurement model assessment Before evaluating our structural model, we must evaluate our measurement model. The assessment of the measurement model guarantees the validity of the measurements and their efficient support of the theoretical components [ 54 ]. To evaluate the measurement model, “construct reliability,” “convergent validity,” and “discriminant validity” are examined [ 33 ]. The construct reliability is assessed by looking at both “composite reliability (CR)” and “Cronbach’s alpha (CA).” For both measurements, the threshold value is 0.7 [ 31 ]. Both the CA and CR values surpass the 0.7 cut-off, as shown in Table 2 , indicating the reliability of the indicators. Convergent validity is assessed using factor loadings (FL) and average variance extracted (AVE). The AVE cut-off value should be above 0.5, and the FL threshold value should be above 0.708 [ 33 ]. Table 2 demonstrates that all FL values are more than 0.708. Furthermore, every AVE value is higher than the 0.5 cut-off. The convergent validity is therefore validated. We assessed discriminant validity using both the Fornell-Larcker criterion and the Heterotrait-Monotrait ratio of correlations (HTMT). Following Fornell and Larcker [ 29 ], discriminant validity is established when each construct’s square root of the AVE exceeds its correlations with other constructs, and this condition was met for all constructs in Table 3 . Additionally, we examined HTMT values, which should remain below 0.85 [ 35 ]. All HTMT values in Table 3 fall below the recommended threshold, further confirming discriminant validity. Table 3. Fornell-Larcker criterion and HTMT results Constructs 1 2 3 4 5 1. AI-induced anxiety 0.906 0.220 0.396 0.201 0.372 2. AI transparency 0.199 0.878 0.259 0.285 0.270 3. Consumer digital literacy 0.368 0.231 0.885 0.841 0.531 4. Consumer satisfaction 0.193 0.260 0.780 0.840 0.275 5. Technostress 0.348 0.237 0.481 0.255 0.885 Open in a new tab Fornell-Larcker criterion (below the diagonal) and HTMT (above the diagonal) Structural model and hypotheses analysis Henseler et al., [ 35 ] proposed the bootstrapping technique with 5000 resamples to assess the structural model and hypothesis analysis. Additionally, Hahn and Ang [ 32 ], argued that researchers shouldn’t base their acceptance or rejection of a hypothesis only on the p-value. They recommended that when analysing hypotheses, the confidence interval (CI) be presented. As a result, CI determined whether a hypothesis was accepted or rejected [ 42 ]. Table 4 ; Fig. 2 display the hypothesis testing results supporting H1 , indicating that AI transparency significantly enhances consumer satisfaction (β = 0.112, p = 0.011). However, both AI-induced anxiety and technostress serve key negative mediating roles in this relationship. The negative mediation effect of AI-induced anxiety (β = -0.018, p = 0.024) suggests that higher AI transparency can sometimes raise customer anxiety, which diminishes satisfaction ( H2 ). Similarly, according to the pronounced negative mediation effect of technostress (β = -0.035, p = 0.002), increased AI transparency may result in technostress, adversely impacting customer satisfaction ( H3 ). Finally, H4 was rejected, indicating that consumer digital literacy did not significantly affect the relationship between AI transparency and consumer satisfaction (β = 0.048, p = 0.415). These findings highlight the complex nature of AI transparency, emphasising the need for careful implementation strategies in e-commerce, as it can both boost satisfaction and induce stress and anxiety. Table 4. Hypothesis testing Relation β-value t-value CI [5.0% − 95.0%] p -value Accepted? H1 : AIT → CSA 0.112 2.545 [0.026–0.197] 0.011 Yes H2 : AIT → AIA → CSA -0.018 2.262 [-0.035 – -0.005] 0.024 Yes H3 : AIT → TST → CSA -0.035 3.162 [-0.059 – -0.016] 0.002 Yes H4 : CDL * AIT → CSA 0.048 0.815 [-0.165–0.067] 0.415 No Open in a new tab Fig. 2. Open in a new tab Results of hypothesis testing Importance-performance map analysis (IPMA) Using the IPMA improves comprehension of PLS-SEM results [ 70 ]. IPMA also considers the average of latent variables and the indicators. The target variable in this study is consumer satisfaction. Figure 3 shows the IPMA results. In terms of importance, the most important variable that affects consumer satisfaction is AI transparency. This is followed by technostress, AI-induced anxiety, and consumer digital literacy. If we examine the performance of each variable, we can observe that all the variables are performing well within the range of 69.639 to 71.051. Fig. 3. Open in a new tab Importance-performance map analysis ANN results It is difficult to understand how AI transparency affects customer satisfaction because PLS-SEM is only good at analysing linear models and correlations; it is unable to handle non-linear interactions. To find and investigate both linear and non-linear correlations, we have used the ANN technique. When compared to traditional regression techniques, using the ANN can produce better predictions [ 4 , 8 ]. However, an ANN technique by itself is insufficient for hypothesis testing because of its “black-box” character [ 19 ], which makes using the ANN in conjunction with the PLS-SEM more successful. We have retrieved the important factors from the PLS-SEM results as inputs to avoid model over-fitting in the ANN [ 3 ]. Three layers comprise the ANN algorithm: input, hidden, and output [ 6 ]. A multi-layer perceptron employing feed-forward-backward propagation was used in ANN analysis. Additionally, to lessen the chance of overfitting, a tenfold cross-validation process was employed [ 7 ]. According to earlier studies, training and testing were conducted using the 90:10 data component [ 6 , 52 , 73 ]. The “root mean square errors” (RMSE) values were then calculated using SPSS, showing that the analysis produced a model fit [ 52 ]. Table 5 presents the RMSE results. Table 5. RMSE results Neural networks Training Testing Total sample N SSE RMSE N SSE RMSE 1 309 2.173 0.083 32 0.154 0.069 341 2 309 2.403 0.088 32 0.204 0.079 341 3 303 2.378 0.088 38 0.510 0.115 341 4 298 3.244 0.104 43 0.293 0.082 341 5 310 2.218 0.084 31 0.153 0.070 341 6 301 2.029 0.082 40 0.389 0.098 341 7 307 2.347 0.087 34 0.184 0.073 341 8 301 1.834 0.078 40 0.603 0.122 341 9 305 1.962 0.080 36 0.394 0.104 341 10 308 2.006 0.081 33 0.449 0.116 341 Mean 2.259 0.085 Mean 0.333 0.092 S. D 0.375 0.006 S. D 0.151 0.019 Open in a new tab Importantly, ANN contributes psychological value beyond statistical prediction. ANN enables the identification of which psychological variables exert the strongest influence on consumer outcomes, even when relationships are non-linear or interaction effects are subtle. This is particularly relevant in affect-driven phenomena such as anxiety and technostress, which often emerge from complex cognitive-emotional processes. Moreover, we acknowledge the interpretational limitations of ANN, as the internal computation processes cannot be directly observed. To address this, we emphasise that ANN findings are used not for causal inference but for uncovering influential psychological predictors that may remain hidden in linear SEM models. This predictive perspective aligns with contemporary psychology, which increasingly recognises the importance of combining explanatory and predictive modelling to understand real-world emotional and behavioural responses to technology. Sensitivity analysis As indicated in Table 6 , sensitivity analysis was performed in the model to evaluate each input neuron’s capacity for prediction. Each neuron’s normalised significance (NI) in the model was calculated based on the analysis and displayed as a percentage [ 77 ]. The sensitivity analysis reveals that consumer digital literacy is the most influential predictor of consumer satisfaction, with the highest average importance (0.725) and NI of 100%. AI transparency ranks second (15.17%), followed closely by technostress (14.92%), while AI-induced anxiety has the lowest influence (7.73%). These results highlight that consumer digital literacy plays a critical role in shaping consumer satisfaction in AI-driven e-commerce environments, whereas AI-induced anxiety has a relatively minor impact. Table 6. Sensitivity analysis Neural network AIA AIT CDL TST 1 0.091 0.134 0.717 0.058 2 0.062 0.073 0.737 0.128 3 0.048 0.130 0.706 0.117 4 0.042 0.078 0.735 0.144 5 0.026 0.130 0.720 0.124 6 0.063 0.126 0.711 0.100 7 0.071 0.052 0.768 0.109 8 0.045 0.119 0.712 0.123 9 0.048 0.130 0.732 0.090 10 0.065 0.124 0.720 0.091 Average importance 0.056 0.109 0.725 0.108 NI (%) 7.73 15.17 100.0 14.92 Ranking 4 2 1 3 Open in a new tab These findings support the psychological relevance of applying ANN: digital literacy emerges as a dominant psychological competence shaping how individuals emotionally interpret and respond to AI-driven interactions. While SEM did not confirm moderation, ANN highlights its central predictive role in understanding consumer reactions. This reinforces the value of integrating predictive models to complement theory-driven psychological explanations. Discussion and conclusions This study investigated how AI transparency influences consumer satisfaction, considering the mediating effects of AI-induced anxiety and technostress, as well as the moderating role of consumer digital literacy, using a hybrid SEM-ANN approach. In addition to its business relevance, the findings offer important psychological insights into how users cognitively and emotionally process AI-driven interactions. First, the positive influence of AI transparency on consumer satisfaction is consistent with earlier work [ 39 , 63 , 66 ]. From a psychological perspective, transparency reduces ambiguity and allows users to form clearer mental models of how AI systems operate, lowering uncertainty and enhancing perceptions of fairness [ 61 ]. According to cognitive load theory [ 83 ], clear system explanations reduce extraneous cognitive demands and lessen the mental effort required to interpret AI recommendations. Transparency may also reduce the need for vigilance, as users feel less compelled to monitor or question algorithmic outcomes. Second, the significant mediating roles of AI-induced anxiety and technostress highlight how transparency can simultaneously alleviate and activate stress-related cognitive processes. AI-induced anxiety differs from general anxiety because it stems specifically from interactions with intelligent systems, such as concerns about algorithmic bias, incorrect decisions, reduced autonomy, and privacy risks [ 18 , 67 ]. This aligns with threat appraisal processes described in the transactional model of stress and coping [ 50 ], where users evaluate potential harm before forming emotional responses. Technostress similarly reflects cognitive overload, rapid system changes, and diminished feelings of competence when dealing with complex AI interfaces [ 56 ]. These findings reveal a paradox: while transparency enhances trust, the increased information about system operations may also heighten awareness of potential risks, contributing to anxiety and mental fatigue [ 69 ]. Third, although SEM results showed that digital literacy does not significantly moderate the relationship between transparency and satisfaction, psychological theory helps explain its deeper influence. Digital literacy serves as a personal coping resource, supporting self-regulation, enhancing perceived control, and reducing stress when interacting with digital systems. Users with higher digital literacy are better equipped to interpret AI outputs, manage cognitive load, and respond constructively to perceived risks [ 22 ]. Although this role did not emerge statistically as a moderator, the ANN analysis identified digital literacy as the most influential predictor of consumer satisfaction, suggesting that its impact may be non-linear and embedded within broader adaptive processes. This interpretation aligns with research indicating that digital competence improves confidence and reduces resistance in AI-based environments [ 13 , 92 ]. The contrasting findings between SEM and ANN reflect the methodological differences between these two approaches. SEM provides theory-driven, linear, and causal inference, while ANN identifies variables that contribute most strongly to predictive accuracy [ 33 , 42 , 82 ]. As the reviewer highlighted, ANN results should not be interpreted causally; they indicate predictive significance rather than causal pathways. The strong predictive influence of digital literacy reinforces its psychological importance, but does not contradict the SEM finding that it does not operate as a traditional statistical moderator. Overall, the findings suggest that AI transparency serves as both a cognitive aid and a psychological stressor. Transparency can reduce uncertainty and cognitive load, leading to higher satisfaction, while simultaneously activating anxiety and technostress through heightened awareness of risks, control loss, or system complexity. Digital literacy plays a crucial role in how effectively consumers manage these demands. Future research could further integrate psychological frameworks, such as cognitive load theory, the transactional stress model, and self-regulation theories, to deepen the understanding of how individuals adapt to transparent AI systems across different contexts. Theoretical implications This study extends current models of AI adoption and consumer behaviour, offering important theoretical implications. By demonstrating that, despite being widely believed to increase trust, AI transparency can also cause anxiety and technostress, which can have conflicting consequences on customer satisfaction, it expands on the TAM. This raises questions about the widely held belief that transparency is an inherently good attribute, suggesting the need for more nuanced approaches to ideas about AI. Furthermore, this study emphasises the psychological toll that AI encounters impose by using the SSO paradigm, highlighting the part that stress-related factors play in influencing consumer experiences. Beyond the conventional emphasis on usability and efficiency, these insights enhance our understanding of how AI-driven systems influence user perceptions. Additionally, the results regarding consumer digital literacy offer fresh theoretical perspectives on how people perceive AI differently. Although the association between AI transparency and satisfaction was not significantly moderated by consumer digital literacy in traditional SEM analysis, its significance in ANN results suggests that digital literacy plays a crucial role in influencing consumer responses in complex and non-linear ways. This emphasizes how useful it is to use hybrid analytical approaches to identify more profound behavioral patterns that traditional statistical methods may miss. This study enhances current theoretical models and lays the groundwork for future research on improving AI-driven customer experiences in e-commerce settings by bridging the gap between digital literacy, psychological stresses, and AI transparency. This study advances the AI transparency literature by reframing transparency not as a uniformly beneficial design principle, but as a psychologically contingent mechanism whose effects depend on users’ cognitive and affective capacities. By jointly modeling AI-induced anxiety and technostress, the findings suggest that transparency simultaneously activates reassurance and cognitive burden, reinforcing the paradoxical nature of transparent AI systems. Accordingly, the study encourages adopting more granular psychological frameworks that differentiate between affective strain, cognitive load, and adaptive capacity in human–AI interaction, thereby extending existing models of AI transparency from static causal explanations toward dynamic, user-centered interpretations. Importantly, this study contributes psychological insights into how individuals cognitively process AI systems, suggesting that transparency-related stress may arise from increased cognitive load, uncertainty appraisal, and heightened attentional vigilance. By linking AI features with underlying mechanisms such as cognitive evaluation, threat perception, and coping resources, the study advances theoretical understanding of how users psychologically adapt to technologically complex environments. Practical implications This study offers several important implications for e-commerce companies employing AI-powered customer service systems. First, the findings indicate that transparency must be carefully designed: while clear explanations can enhance trust, overly technical or lengthy disclosures may increase AI-induced anxiety and technostress. Therefore, companies should adopt cognitively light transparency strategies by providing concise, user-friendly explanations of AI decisions and avoiding unnecessary technical detail. To translate these insights into practical actions, the IPMA results can be converted into a simple managerial checklist. Firms should prioritize: (1) concise explanations, offer short, plain-language reasons for AI outputs; (2) cognitive-load reduction, simplify interface design, avoid jargon, and present information progressively; (3) human-assist options, allow seamless escalation to a human agent when users experience confusion or distress; (4) digital literacy support, provide tutorials, interactive onboarding, and adaptive interfaces for users with varying skill levels. Strengthening consumer digital literacy is particularly important because it shapes how consumers psychologically process AI interactions. Providing training resources, personalized assistance, or interface customization can help reduce anxiety, improve user adjustment, and enhance overall psychological comfort. By implementing these strategies, e-commerce platforms can reduce the negative emotional consequences of transparency while improving user satisfaction, trust, and long-term engagement. From a psychological perspective, the findings carry substantial implications for everyday technological stress and well-being. AI-induced anxiety and technostress may disrupt emotional regulation, increase cognitive fatigue, and diminish users’ sense of control. Designers and practitioners should consider incorporating features that promote psychological comfort, such as reassurance cues, adaptive transparency levels, and options for human support, to mitigate anxiety and support healthier technology adjustment. The study also points to the possibility of psychological interventions. Digital well-being training, stress-coping resources, and user education programs could help individuals develop more adaptive coping mechanisms when interacting with AI systems. Such interventions can enhance emotional resilience, reduce uncertainty-driven stress, and foster more secure engagement with AI technologies. Ultimately, the findings provide practitioners with insight into the cognitive mechanisms underlying AI perception. By reducing cognitive load, clarifying decision processes, and enhancing perceived controllability, organizations can facilitate smoother, psychologically informed user experiences that support both satisfaction and emotional well-being. Psychological contribution This study advances psychological theory by illuminating the dual-edged nature of AI transparency in shaping consumer cognition and affect. While transparency is commonly framed as a trust-enhancing mechanism, the findings reveal that increased transparency can simultaneously activate AI-induced anxiety and technostress, introducing psychological costs that undermine consumer satisfaction. By positioning these constructs as parallel mediators, the study extends technostress and technology anxiety literature into AI-driven service contexts, demonstrating that transparency does not operate as a purely positive cognitive cue but also triggers affective strain and uncertainty. This understanding contributes to consumer psychology by showing how heightened awareness of AI decision-making processes can overload consumers’ cognitive resources, intensify perceived loss of control, and evoke stress-related responses during human–AI interactions. Furthermore, the study contributes to psychological methodology and individual-difference research by integrating SEM-ANN to uncover both linear causal mechanisms and non-linear psychological salience. Although consumer digital literacy does not function as a significant moderator in the SEM model, ANN results identify it as the most influential predictor of consumer satisfaction, suggesting a threshold-based or non-linear psychological effect. This highlights the limitation of traditional linear models in capturing complex cognitive–emotional processes underlying AI use. Psychologically, the findings reposition digital literacy as a core self-regulatory resource that shapes how consumers interpret, cope with, and emotionally respond to transparent AI systems. By mixing cognitive appraisal theory, technostress theory, and hybrid analytical modeling, this study enhances our understanding of how consumers psychologically process AI transparency and adapt to increasingly autonomous digital environments. Limitations and future research directions Several limitations should be acknowledged. First, the sample was collected exclusively from Chinese e-commerce consumers, which may limit the generalisability of the findings to other cultural, regional, and regulatory contexts. Future studies should therefore conduct multi-region or cross-cultural comparisons to determine whether psychological responses to AI transparency differ across markets. Second, the study employed a cross-sectional design, preventing strong causal inference. Future research should utilise longitudinal approaches to track how AI-induced anxiety, technostress, and satisfaction evolve over repeated interactions. Additionally, experimental designs, such as manipulating levels or types of transparency, would allow for clearer causal conclusions about how specific transparency elements regulate emotional and cognitive responses. Third, although this research used a hybrid SEM-ANN approach, it did not examine contextual factors such as industry differences, service complexity, or variation in AI capability. Future work could compare AI applications across multiple sectors or levels of algorithmic sophistication. Moreover, the increasing role of generative AI (e.g., large language model–enhanced chatbots) presents an opportunity for deeper investigation into emotional regulation, perceived autonomy, and psychological adjustment in human-AI interactions. Ultimately, future research may incorporate additional psychological constructs, such as perceived control, risk appraisal, or coping strategies, to provide a more comprehensive understanding of the mechanisms by which transparency influences user well-being and satisfaction. Lastly, this study relies on purposive online sampling, which may limit the sample’s representativeness and constrain the external validity of the findings. Moreover, the absence of explicit consideration of cultural, demographic, and contextual heterogeneity restricts the generalizability of the results beyond the specific group of Chinese e-commerce consumers examined. Supplementary Information Supplementary Material 1. (16.4KB, docx) Authors’ contributions M.F.S. conceptualization, data curation, original draft. M.A. software, analysis, methodology. A.M.M. editing, supervision. W.J. editing, revision. C.H. methodology, editing, revision. Funding Funding: This research was supported by the following funded projects: the 2025 Zhejiang Provincial Philosophy and Social Sciences Planning Project, Digitalization and Multi-Scenario Dissemination of Hemudu Culture from the Perspective of the Chinese National Spirit Identity (Grant No. 25NDJC163YB); and the 2024 Ningbo Municipal First-Batch Philosophy and Social Sciences Planning Project, Decoding the Cultural Genes of Hemudu Culture and Their Scenario-Based Applications in the Context of Cultural Data Elements (Grant No. G2024-1-69). Data availability Data will be made available on reasonable demand. Declarations Ethical approval and consent to participate All participants provided informed consent before taking part in the survey. They were assured of voluntary participation, confidentiality, and the use of data solely for research purposes. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of the University of the Punjab, Lahore, Pakistan. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Abrardi L, Cambini C, Rondi L. Artificial intelligence, firms and consumer behavior: A survey. J Economic Surveys. 2022;36(4):969–91. [ Google Scholar ] 2. Addai P, Delle E, Okyere I, Amponsah R, Govina SE. Disrupting cognitive capacity through excessive technology use among academic staff: examining technostress, ethical decision-making, emotional exhaustion, and ethical climate. SN Bus Econ. 2025;5(10):149. [ Google Scholar ] 3. Al-Emran M, Mezhuyev V, Kamaludin A. Towards a conceptual model for examining the impact of knowledge management factors on mobile learning acceptance. Technology in Society. 2020. 10.1016/j.techsoc.2020.101247. 4. Al-Sharafi MA, Iranmanesh M, Al-Emran M, Alzahrani AI, Herzallah F, Jamil N. Determinants of cloud computing integration and its impact on sustainable performance in SMEs: anempirical investigation using the SEM-ANN approach. Heliyon. 2023;9(5):e16299. 10.1016/j.heliyon.2023.e16299. [ DOI ] [ PMC free article ] [ PubMed ] 5. Alam MMD, Alam MZ, Rahman SA, Taghizadeh SK. Factors influencing mHealth adoption and its impact on mental well-being during COVID-19 pandemic: A SEM-ANN approach. J Biomed Inform. 2021;116:103722. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Ali K, Johl SK, Muneer A, Alwadain A, Ali RF. Soft and hard total quality management practices promote industry 4.0 readiness: a SEM-neural network approach. Sustainability. 2022;14(19):11917. [ Google Scholar ] 7. Ali RF, Jahankhani H, Ali K, Hassan B. Blockchain-Based security factors on sustainable supply chain management in UK manufacturing firms: A hybrid SEM-ANN approach. Systems. 2024;12(6):208. [ Google Scholar ] 8. Alkawsi GA, Ali N, Mustafa AS, Baashar Y, Alhussian H, Alkahtani A, Tiong SK, Ekanayake J. A hybrid SEM-neural network method for identifying acceptance factors of the smart meters in malaysia: challenges perspective. Alexandria Eng J. 2021;60(1):227–40. [ Google Scholar ] 9. Alowais SA, Alghamdi SS, Alsuhebany N, Alqahtani T, Alshaya AI, Almohareb SN, Aldairem A, Alrashed M, Bin Saleh K, Badreldin HA. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ. 2023;23(1):689. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Apell P, Eriksson H. Artificial intelligence (AI) healthcare technology innovations: the current state and challenges from a life science industry perspective. Technol Anal Strateg Manag. 2023;35(2):179–93. [ Google Scholar ] 11. Arpaci I. A multi-analytical SEM-ANN approach to investigate the social sustainability of AI chatbots based on cybersecurity and protection motivation theory. IEEE Trans Eng Manage. 2023;71:1714–25. [ Google Scholar ] 12. Asif M, Fazel H. (2024). Factors influencing tourists’ technology adoption in Saudi Arabia: examining determinants for effective use of mobile apps in tourism. Global Knowledge, Memory and Communication , ahead-of http://-print(ahead-of-print) 13. Asif M, Sarwar F. Customer relationship management, financial literacy, and online banking adoption: social influence moderation-insights from technology adoption. RADS J Bus Manage. 2023;5(2):129–44. [ Google Scholar ] 14. Asif M, Sarwar F. Impact of customer relationship management, financial literacy and social influence on online banking adoption: the moderating role of personal innovativeness. Asia-Pacific J Bus Adm. 2024;17(3):744–64. [ Google Scholar ] 15. Baek TH, Kim M. Is ChatGPT scary good? How user motivations affect creepiness and trust in generative artificial intelligence. Telematics Inform. 2023;83:102030. [ Google Scholar ] 16. Bedué P, Fritzsche A. Can we trust AI? An empirical investigation of trust requirements and guide to successful AI adoption. J Enterp Inform Manage. 2022;35(2):530–49. [ Google Scholar ] 17. Campbell S, Greenwood M, Prior S, Shearer T, Walkem K, Young S, Bywaters D, Walker K. Purposive sampling: complex or simple? Research case examples. J Res Nurs. 2020;25(8):652–61. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Chen C, Hu W, Wei X. From anxiety to action: exploring the impact of artificial intelligence anxiety and artificial intelligence self-efficacy on motivated learning of undergraduate students. Interact Learn Environ. 2025;33(4):3162–77. [ Google Scholar ] 19. Chong AY-L. A two-staged SEM-neural network approach for Understanding and predicting the determinants of m-commerce adoption. Expert Syst Appl. 2013;40(4):1240–7. [ Google Scholar ] 20. Daly SJ, Hearn G, Papageorgiou K. Sensemaking with AI: how trust influences human-AI collaboration in health and creative industries. Social Sci Humanit Open. 2025;11:101346. [ Google Scholar ] 21. Daud NM. From innovation to stress: analyzing hybrid technology adoption and its role in technostress among students. Int J Educational Technol High Educ. 2025;22(1):31. [ Google Scholar ] 22. Donmez O, Seferoğlu SS. The role of ICT Self-Efficacy, Technostress, individual innovativeness and mindset on artificial intelligence literacy: artificial intelligence literacy. Int J Curriculum Instruction. 2025;17(3):950–77. [ Google Scholar ] 23. Duong CD. Modeling the determinants of HEI students’ continuance intention to use ChatGPT for learning: a stimulus–organism–response approach. J Res Innovative Teach Learn. 2024;17(2):391–407. [ Google Scholar ] 24. Eppler MJ, Mengis J. The concept of information overload: A review of literature from organization science, accounting, marketing, MIS, and related disciplines. Inform Soc. 2004;20(5):325–44. [ Google Scholar ] 25. Faul F, Erdfelder E, Buchner A, Lang A-G. Statistical power analyses using G* power 3.1: tests for correlation and regression analyses. Behav Res Methods. 2009;41(4):1149–60. [ DOI ] [ PubMed ] [ Google Scholar ] 26. Felzmann H, Villaronga EF, Lutz C, Tamò-Larrieux A. Transparency you can trust: transparency requirements for artificial intelligence between legal norms and contextual concerns. Big Data Soc. 2019;6(1):2053951719860542. [ Google Scholar ] 27. Li Z, Zhang W, Zhang H, Gao R, Fang X. Global digital compact: a mechanism for the governance of online discriminatory and misleading content generation. Int J Hum Comput Interact. 2025;41(2):1381–96. 10.1080/10447318.2024.2314350. 28. Flavián C, Pérez-Rueda A, Belanche D, Casaló LV. Intention to use analytical artificial intelligence (AI) in services–the effect of technology readiness and awareness. J Service Manage. 2022;33(2):293–320. [ Google Scholar ] 29. Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39–50. [ Google Scholar ] 30. Frank D-A, Chrysochou P, Mitkidis P, Otterbring T, Ariely D. Navigating uncertainty: exploring consumer acceptance of artificial intelligence under self-threats and high-stakes decisions. Technol Soc. 2024;79:102732. [ Google Scholar ] 31. Gefen D, Straub D, Boudreau M-C. Structural equation modeling and regression: guidelines for research practice. Commun Association Inform Syst. 2000;4(1):1–77. [ Google Scholar ] 32. Hahn ED, Ang SH. From the editors: new directions in the reporting of statistical results in the. J World Bus. 2017;52(2):125–6. [ Google Scholar ] 33. Hair JF, Sarstedt M, Ringle CM. Rethinking some of the rethinking of partial least squares. Eur J Mark. 2019;53(4):566–84. [ Google Scholar ] 34. Haleem A, Javaid M, Qadri MA, Singh RP, Suman R. Artificial intelligence (AI) applications for marketing: A literature-based study. Int J Intell Networks. 2022;3:119–32. [ Google Scholar ] 35. Henseler J, Ringle CM, Sarstedt M. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J Acad Mark Sci. 2015;43:115–35. [ Google Scholar ] 36. Hentzen JK, Hoffmann A, Dolan R, Pala E. Artificial intelligence in customer-facing financial services: a systematic literature review and agenda for future research. Int J Bank Mark. 2022;40(6):1299–336. [ Google Scholar ] 37. Hornung O, Smolnik S. AI invading the workplace: negative emotions towards the organizational use of personal virtual assistants. Electron Markets. 2022;32(1):123–38. [ Google Scholar ] 38. Hossain MA, Agnihotri R, Rushan MRI, Rahman MS, Sumi SF. Marketing analytics capability, artificial intelligence adoption, and firms’ competitive advantage: evidence from the manufacturing industry. Ind Mark Manage. 2022;106:240–55. [ Google Scholar ] 39. Hu P, Zeng Y, Wang D, Teng H. Too much light blinds: the transparency-resistance paradox in algorithmic management. Comput Hum Behav. 2024;161:108403. [ Google Scholar ] 40. Hussain I, Sabir MR, ur Rehman N, Ghaffar I, Majeed KB. A Spatial of digital Technology, digital Literacy, performance expectancy and techno stress in pandemic conditions in technological institutes. J Disaster Recovery Bus Continuity. 2022;13(1):140–9. [ Google Scholar ] 41. Issa H, Jaber J, Lakkis H. Navigating AI unpredictability: exploring technostress in AI-powered healthcare systems. Technol Forecast Soc Chang. 2024;202:123311. [ Google Scholar ] 42. Johl SK, Ali K, Shirahada K, Oyewale OI. Green servitization, circular economy, and sustainability a winning combination analysis through hybrid SEM‐ANN approach. Bus Strategy Environ. 2024;33(8):8978–93. 43. Kim SY, Park H, Kim H, Kim J, Seo K. Technostress causes cognitive overload in high-stress people: eye tracking analysis in a virtual kiosk test. Inf Process Manag. 2022;59(6):103093. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Kock N. Common method bias in PLS-SEM: A full collinearity assessment approach. Int J e-Collaboration (ijec). 2015;11(4):1–10. [ Google Scholar ] 45. Kock N, Lynn GS. Lateral collinearity and misleading results in variance-based SEM: an illustration and recommendations. J Association Inform Syst. 2012;13(7):546–80. [ Google Scholar ] 46. Kohnke L, Zou D, Moorhouse BL. Technostress and english Language teaching in the age of generative AI. Educational Technol Soc. 2024;27(2):306–20. [ Google Scholar ] 47. Kulal A, Rahiman HU, Suvarna H, Abhishek N, Dinesh S. Enhancing public service delivery efficiency: exploring the impact of AI. J Open Innovation: Technol Market Complex. 2024;10(3):100329. [ Google Scholar ] 48. Kumar V, Rajan B, Salunkhe U, Joag SG. Relating the dark side of new-age technologies and customer technostress. Psychol Mark. 2022;39(12):2240–59. [ Google Scholar ] 49. Larbi-Siaw O, Xuhua H, Owusu E, Owusu-Agyeman A, Fulgence BE, Frimpong SA. Eco-innovation, sustainable business performance and market turbulence moderation in emerging economies. Technol Soc. 2022;68:101899. [ Google Scholar ] 50. Lazarus RS. Stress, appraisal, and coping. Volume 445. Springer; 1984. 51. Lehmann CA, Haubitz CB, Fügener A, Thonemann UW. The risk of algorithm transparency: how algorithm complexity drives the effects on the use of advice. Prod Oper Manage. 2022;31(9):3419–34. [ Google Scholar ] 52. Leong L-Y, Hew J-J, Lee V-H, Tan GW-H, Ooi K-B, Rana NP. An SEM-ANN analysis of the impacts of blockchain on competitive advantage. Industrial Manage Data Syst. 2023;123(3):967–1004. [ Google Scholar ] 53. Li X, Liu Z, Chen Y, Ren A. Consumer avoidance toward message stream advertising on mobile social media: a stimulus-organism-response perspective. Inform Technol People. 2025;38(1):23–47. [ Google Scholar ] 54. Lowry PB, Gaskin J. Partial least squares (PLS) structural equation modeling (SEM) for Building and testing behavioral causal theory: when to choose it and how to use it. IEEE Trans Prof Commun. 2014;57(2):123–46. [ Google Scholar ] 55. Lu W. Inevitable challenges of autonomy: ethical concerns in personalized algorithmic decision-making. Humanit Social Sci Commun. 2024;11(1):1–9. [ Google Scholar ] 56. Maier C, Laumer S, Thatcher JB, Wirth J, Weitzel T. Trial-period technostress: A conceptual definition and mixed-methods investigation. Inform Syst Res. 2022;33(2):489–514. [ Google Scholar ] 57. Malhotra G, Kharub M. Elevating logistics performance: Harnessing the power of artificial intelligence in e-commerce. Int J Logistics Manage. 2025;36(1):290–321. [ Google Scholar ] 58. Mehrabian A, Russell JA. The basic emotional impact of environments. Percept Mot Skills. 1974;38(1):283–301. [ DOI ] [ PubMed ] [ Google Scholar ] 59. Milošević I, Ruso J, Glogovac M, Arsić S, Rakić A. An integrated SEM-ANN approach for predicting QMS achievements in industry 4.0. Total Qual Manage Bus Excellence. 2022;33(15–16):1896–912. [ Google Scholar ] 60. Mogaji E, Jain V. How generative AI is (will) change consumer behaviour: postulating the potential impact and implications for research, practice, and policy. J Consumer Behav. 2024;23(5):2379–89. [ Google Scholar ] 61. Monod E, Mayer A-S, Straub D, Joyce E, Qi J. From worker empowerment to managerial control: the devolution of AI tools’ intended positive implementation to their negative consequences. Inf Organ. 2024;34(1):100498. [ Google Scholar ] 62. Murtas G, Pedeliento G. Investigating the customer journey in Second-Hand fashion platforms: implications for luxury brand management. J Consumer Behav. 2025;24(2):655–72. [ Google Scholar ] 63. Nizette F, Hammedi W, van Riel AC, Steils N. Why should I trust you? Influence of explanation design on consumer behavior in AI-based services. J Service Manage. 2025;36(1):50–74. [ Google Scholar ] 64. Okeleke PA, Ajiga D, Folorunsho SO, Ezeigweneme C. Predictive analytics for market trends using AI: A study in consumer behavior. Int J Eng Res Updates. 2024;7(1):36–49. [ Google Scholar ] 65. Omoge AP, Gala P, Horky A. Disruptive technology and AI in the banking industry of an emerging market. Int J Bank Mark. 2022;40(6):1217–47. [ Google Scholar ] 66. Park K, Yoon HY. Beyond the code: the impact of AI algorithm transparency signaling on user trust and relational satisfaction. Public Relations Rev. 2024;50(5):102507. [ Google Scholar ] 67. Peng Y, Wang Y, Li J, Yang Q. Impact of AI-oriented live-streaming E-commerce service failures on consumer disengagement—empirical evidence from China. J Theoretical Appl Electron Commer Res. 2024;19(2):1580–98. [ Google Scholar ] 68. Podsakoff PM, MacKenzie SB, Lee J-Y, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. J Appl Psychol. 2003;88(5):879–903. [ DOI ] [ PubMed ] [ Google Scholar ] 69. Priporas CV, Hussain S, Khaneja S, Rahman H. Technology distraction in generation Z: the effects on consumer responses, sensory overload, and discomfort. Int J Inf Manag. 2024;75:102751. [ Google Scholar ] 70. Ringle CM, Sarstedt M. Gain more insight from your PLS-SEM results: the importance-performance map analysis. Industrial Manage Data Syst. 2016;116(9):1865–86. [ Google Scholar ] 71. Sanchez TW, Brenman M, Ye X. The ethical concerns of artificial intelligence in urban planning. J Am Plann Association. 2024;91(2):294–307. [ Google Scholar ] 72. Shahzad F, Xiu G, Khan MAS, Shahbaz M. Predicting the adoption of a mobile government security response system from the user’s perspective: an application of the artificial neural network approach. Technol Soc. 2020;62:101278. [ Google Scholar ] 73. Shahzad MF, Xu S, An X, Zahid H, Asif M. Learning and teaching in the era of generative artificial intelligence technologies: an In-Depth exploration using Multi‐Analytical SEM‐ANN approach. Eur J Educ. 2025;60(1):e70050. 10.1111/ejed.70050. 74. Sharma S, Gupta B. Investigating the role of technostress, cognitive appraisal and coping strategies on students’ learning performance in higher education: a multidimensional transactional theory of stress approach. Inform Technol People. 2023;36(2):626–60. [ Google Scholar ] 75. Shiau W-L, Huang L-C. Scale development for analyzing the fit of real and virtual world integration: an example of Pokémon go. Inform Technol People. 2023;36(2):500–31. [ Google Scholar ] 76. Shin D, Kee KF, Shin EY. Algorithm awareness: why user awareness is critical for personal privacy in the adoption of algorithmic platforms? Int J Inf Manag. 2022;65:102494. [ Google Scholar ] 77. Siddik AB, Yong L, Rahman MN. The role of fintech in circular economy practices to improve sustainability performance: a two-staged SEM-ANN approach. Environ Sci Pollut Res. 2023;30(49):107465–86. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 78. Singh P, Singh V. The power of AI: enhancing customer loyalty through satisfaction and efficiency. Cogent Bus Manage. 2024;11(1):2326107. [ Google Scholar ] 79. Sison AJG, Daza MT, Gozalo-Brizuela R, Garrido-Merchán EC. ChatGPT: more than a weapon of mass deception ethical challenges and responses from the human-centered artificial intelligence (HCAI) perspective. Int J Human–Computer Interact. 2024;40(17):4853–72. [ Google Scholar ] 80. Song L, Yao S, Liu L, Tso G. Investigating binge-watching and its effect on paid subscription: A mixed‐method study based on SOR theory. J Consumer Behav. 2025;24(1):20–43. [ Google Scholar ] 81. Suh A, Prophet J. The state of immersive technology research: A literature analysis. Comput Hum Behav. 2018;86:77–90. [ Google Scholar ] 82. Suhail F, Adel M, Al-Emran M, AlQudah AA. Are students ready for robots in higher education? Examining the adoption of robots by integrating UTAUT2 and TTF using a hybrid SEM-ANN approach. Technol Soc. 2024;77:102524. [ Google Scholar ] 83. Sweller J. Cognitive load during problem solving: effects on learning. Cogn Sci. 1988;12(2):257–85. [ Google Scholar ] 84. Tong H, Yu J, Shou M. Mitigating the effect of AI anxiety on employees’ creativity: A social cognitive perspective. J Digit Manage. 2025;1(1):7. [ Google Scholar ] 85. Wang Y-Y, Wang Y-S. Development and validation of an artificial intelligence anxiety scale: an initial application in predicting motivated learning behavior. Interact Learn Environ. 2022;30(4):619–34. [ Google Scholar ] 86. Wanner J, Herm L-V, Heinrich K, Janiesch C. The effect of transparency and trust on intelligent system acceptance: evidence from a user-based study. Electron Markets. 2022;32(4):2079–102. [ Google Scholar ] 87. Willems J, Schmid MJ, Vanderelst D, Vogel D, Ebinger F. AI-driven public services and the privacy paradox: do citizens really care about their privacy? Public Manage Rev. 2023;25(11):2116–34. [ Google Scholar ] 88. Wu R, Ishfaq K, Hussain S, Asmi F, Siddiquei AN, Anwar MA. Investigating e-retailers’ intentions to adopt cryptocurrency considering the mediation of technostress and technology involvement. Sustainability. 2022;14(2):641. [ Google Scholar ] 89. Xu X-Y, Jia Q-D, Tayyab SMU. Exploring the stimulating role of augmented reality features in E-commerce: A three-staged hybrid approach. J Retailing Consumer Serv. 2024;77:103682. [ Google Scholar ] 90. Yang H, Li D, Hu P. Decoding algorithm fatigue: the role of algorithmic literacy, information cocoons, and algorithmic opacity. Technol Soc. 2024;79:102749. [ Google Scholar ] 91. Zhang Q, Xiong Y. Harnessing AI potential in E-Commerce: improving user engagement and sales through deep learning-based product recommendations. Curr Psychol. 2024;43(38):30379–401. [ Google Scholar ] 92. Zhang J, Zhang Z. AI in teacher education: unlocking new dimensions in teaching support, inclusive learning, and digital literacy. J Comput Assist Learn. 2024;40(4):1871–85. [ Google Scholar ] 93. Zhang Y, Zhu J, Chen H, Jiang Y. Enhancing trust and empathy in marketing: Strategic AI and human influencer selection for optimized content persuasion. J Consum Behav. 2025;24(2):866–85. 10.1002/cb.2423. 94. Zhao W, Tang D, Mehta AM, Yu Y, Shahzad MF, Asif M. Adoption of green technologies, smart energy use, and public environmental awareness: A SEM‐ANN model for sustainable energy transition. Environmental Progress & Sustainable Energy. 2025:e70291. 10.1002/ep.70291. 95. Zhu L, Li H, Wang F-K, He W, Tian Z. How online reviews affect purchase intention: a new model based on the stimulus-organism-response (S-O-R) framework. Aslib J Inform Manage. 2020;72(4):463–88. [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Material 1. (16.4KB, docx) Data Availability Statement Data will be made available on reasonable demand. Articles from BMC Psychology are provided here courtesy of BMC ACTIONS View on publisher site PDF (2.2 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top