Innovation resistance theory (IRT) offers an inclusive framework for understanding the user resistance behavior to innovations (Ram and Sheth, 1989). According to IRT, individuals may resist or feel hesitant to accept innovation that they perceive to be risky, irrelevant, or contradictory to the existing status quo and pre-existing value system (Hew et al., 2019). The resistance by the users plays a crucial role in determining whether the innovation will be a success or failure (Ram and Sheth, 1989; Kaur et al., 2020). An investor’s resistance may vary from active to passive resistance (Chawla et al., 2024a; Chawla et al., 2024b). Active resistance stems from the issues related to innovation’s perceived utility, value, and perceived usefulness, whereas passive resistance stems from the psychological barriers such as traditional ideas and beliefs. IRT’s comprehensive approach makes it idyllic for the evaluation of investors’ resistance to innovation (Chawla et al., 2024a; Chawla et al., 2024b). This IRT framework differs from other frameworks as it primarily focuses on value, tradition, usage, risk, and image (Gupta and Arora, 2017). Unlike the models such as UTAUT and TAM, which have a majority of emphasis on the technology adoption through the constructs of ease of use, perceived usefulness etc., the innovation resistance theory (IRT) provides a better perspective of directly addressing the challenges related to the functional and psychological resistance that impede the technology adoption. The distinction becomes quite significant specifically in the context of FRAs, where the reluctance to adopt them is not only influenced by the performance expectations but also by the inertia, behavioral tendencies, and other perceived uncertainties which are often overlooked by TAM and UTAUT. Similarly, although the behavioral reasoning theory (BRT) also acknowledges the conceptual beliefs and the motivations behind the user decision-making, still it does not specifically dissect the innovation-related factors as the theory of innovation resistance does through its structural approach with special focus on functional and psychological resistance. Although the aim of the present research is to explore the reluctance in both emotional and cognitive responses to AI-based robo-advisors, IRT is theoretically more aligned. This clarification will strengthen the study’s conceptual foundation and also validate the selection of the IRT model in the research study.

The literature indicates an increasing interest in understanding innovation resistance, particularly in context to the digitalized services. Numerous studies have employed IRT theoretical framework such as online gamification (Oktavianus et al., 2017), banking services (Matsuo et al., 2018), online travel (Talwar et al., 2020), mobile banking (Kaur et al., 2020), online communities (Kumar et al., 2025), and online-to-offline (O2O) technology platform (Chawla et al., 2024a; Chawla et al., 2024b).

Unlike other online platforms, robo-advisory services involve sensitive financial data, complex algorithms procedures, and financial implications for the users (D’Acunto and Rossi, 2021). There is also a distinction between the user–advisor interactions, lacking the nuanced knowledge and emotional intelligence associated with the human advisors. The existing studies focus on the technology adoption factors, ignoring the potential resistance factors arising from these unique attributes. As a result, our study applies the IRT framework to investigate the factors influencing the resistance behavior of the investors in the robo-advisory services context. Although the previous research studies focusing on the IRT have largely examined the consumer barriers and the adoption intention relationships such as tourism sector (Ahmad and Rasheed, 2025); context to Mobile Payments Systems (Kaur et al., 2020), considerably very less attention has been paid to user intentions and intention to recommend, i.e., the adoption to recommend. Notably, there is very scarce research on the empirical investigations understanding the resistance barriers that to on the adoption outcomes in the domain of robo-advisors specifically (Cardillo and Chiappini, 2024). Thereby, by inculcating the use intention and the intention to recommend within the IRT framework would bridge this gap and extent the IRT model of innovation resistance toward the adoption patterns. The integration of these constructs would capture both the resistance and the enabling mechanisms, i.e., intention to use and intention to recommend, enhancing the explanatory power of the conceptual model.

Attitude toward AI has emerged as a pivotal construct that has a major role in shaping individuals’ acceptance or reluctance to AI-driven technologies such as financial robo-advisors (FRAs). The theory of innovation resistance underscores that resistance by the individuals is not a sole function of the perceived barriers but it is also impacted by their beliefs and pre-existing beliefs (Ram and Sheth, 1989). A positive attitude toward the AI can work as a cognitive filter that changes the perceptions related to risks and barriers, further transforming the perceived potential threats into opportunities for better decision-making (Araujo et al., 2020). The research on AI highlights that the positive attitudes can alter the belief system, mitigating privacy concerns and can reduce perceived complexities barriers (Oprea et al., 2024), but, at the same time, it is also affected by their pre-existing beliefs and attitudes (Ram and Sheth, 1989).

In addition, the users with a positive stance on AI are less susceptible to psychological biases such as inertia and overconfidence as they feel more comfortable in delegating their tasks to the intelligent systems (Longoni et al., 2019). Overall, a positive attitude toward AI enhances the technological readiness, which further helps in lowering down the psychological differences between the users and technology systems, thus reducing the resistance (Parasuraman, 2000). In context to algorithm-driven FRAs, the decision-making process is data-driven and the attitude toward using AI can realign perceptions, reducing resistance and fostering their engagement. The integration of this moderating variable in the model will extend the IRT by demonstrating how the user orientations toward AI can condition the strengths of both psychological barrier and functional barriers that shape the pathways to the adoption of digital financial systems.

Investor resistance is most often driven by the usage barrier. The barrier arises when the innovation does not align with current workflows, procedures, and habits to hold new systems for their advantage. In other words, the sophisticated algorithm-based FRAs are significantly affected by the users’ perceptions of their complexity. According to Chuah et al. (2021), complexity is defined as the situation in which an innovation is perceived to be difficult to understand and use. Complexity can be further divided into (a) complexity of innovative idea (understandability) and (b) complexity of executing idea (usage) (Ram and Sheth, 1989). Some individuals are still fearful or scared to use any new disruptive technology and imagine as something frightening just like a monster (Chuah et al., 2021). A study by Belanche et al. (2019) has suggested that individual’s behavior toward the new robo-based technologies is quite complex and needs consideration both the designing and the traits of the individuals. This complexity can further lead to user resistance to robo-advisory services. In context of FRAs, investors may feel significant degrees of complexities in terms of understanding and effectively utilizing them. This might increase their strain and lead to resistance, resulting in higher stress levels. Studies have shown that whenever an individual’s cognitive load increases, it might lead to negative emotions and reluctance to engage with AI-based FRAs. However, an individual attitude toward AI might influence the extent to which the perceived complexity drives resistance toward FRAs. When discussing AI-based technology, Oprea et al. (2024) support the notion that a user with a more positive attitude toward AI would consider its intricacies inconsequential and vice versa. Thus, the moderating effort of AI between the perceived complexity and the resistance toward FRA also needs to be evaluated. Therefore, it proposes the following hypotheses:

The term ‘value barrier’ denotes the reluctance toward an innovation due to lack of alignment with the benefits and existing value, specifically in balancing costs and perceived benefits (Lyu et al., 2024). To make FRAs appealing, the potential investors must recognize its value (Bhatia et al., 2021). In other words, without apparent value, the resistance toward innovative services would be a natural response (Ram and Sheth, 1989). When the perceived costs exceed the perceived benefits, then the value barrier arises. In context of FRAs, potential investors may resist the use of these platforms, if they doubt the algorithm transparency, or other personalized insights which are previously offered by the traditional human advisors. Increased perceived costs over its perceived benefits is one of the major reasons that hinders the adoption of an innovative service as stated in a study by Kaur et al. (2020). Although FRAs provide numerous benefits, they are unable to answer the queries of the potential users about platforms ability to deliver worthy financial outcomes raising concerns and psychological resistance by the investors. However, the degree of the resistance may not be same/uniform among all investors as their attitude toward AI could play a crucial role in influencing this relationship. The investors with a positive attitude and mindset toward AI may be more willing to re-consider the value provided by FRAs, thus reducing the impact of value barriers on resistance. On the contrary, those who are skeptical or afraid of AI may suffer a heighten resistance, even if the value propositions are strengthen. Thus, building on the following, the study proposes the following hypotheses:

Data privacy risk is an individual’s anxiety about the potential threats about their personal information while utilizing a certain system or service. This anxiety stems from the fear that any unauthorized use of sensitive data may lead to harm or misuse. Data privacy risks might lead to negative emotions on the part of the individual doubting the ability of the technology-based services. Numerous studies have done research on privacy risk factors such as facial payment recognition systems (Liu et al., 2021), home IOT systems (Lee, 2020), and smart services (Mani and Chouk, 2022). In context to information system studies, data privacy risks are one the most vital factor that acts as a barrier in accepting automated-based technologies such as FRAs. Thus, the IRT model considers data privacy risk as a very vital barrier to use any innovative technology. In context of FRAs, they gather sensitive financial information, which makes them vulnerable to data breaches. The susceptibility of these digital platforms to data breaches may elevate investor concerns about the safety of their personal information leading to cognitive stress.

However, the investors with a difference in attitude toward AI can build up these perceptions significantly. Sindermann et al. (2021) suggest that individuals holding a strong and favorable attitude in AI reliability and efficiency can actually buffer the negative impacts of the perceived privacy concerns and vice-versa. Thereby, attitude toward the AI might play a moderating role in softening their resistance to use FRAs. Building on this, the study proposes a hypotheses:

Overconfidence bias refers to a situation where individuals tend to overrate their knowledge, control over financial results, or predictive abilities, leading to cognitive distortion (Zheng et al., 2025). In other words, it is the difference between the subjective beliefs of an individual and objectively measurable outcomes (Piehlmaier, 2022). Potential investors who are overconfident prefer to depend on their own judgment, underrating the potential value of professional or algorithm-based management. The outlook to perceive that algo-based FRAs is inferior or irrelevant when compared with investors own knowledge and judgement generates psychological resistance to delegate sensitive financial decisions to automated devices, that to with minimal human interference. This further might leads to distrust in algorithm-based recommendations, when compared to their own set of intuitions. Previous studies also suggest that confidence actually influences the adoption and spread of innovation and vice-versa (Karki et al., 2024; Germann and Merkle, 2023; Piehlmaier, 2022). However, it is also a matter of consideration that not all of the investors perceive AI uniformly. Those investors with positive and more favorable attitude and confidence toward AI-based technologies may become more open to delegate their tasks to intelligent automated systems, even though if initially they had exhibited overconfidence. A study by Pal et al. (2025) also suggests that a favorable orientation toward using AI can help in reducing the algorithm aversion and vice-versa. This indicates AAI may moderate the relationship strength between overconfidence bias and resistance to FRAs. Therefore, building on this prospect, the following hypotheses are formed:

Image barrier refers to the negative impression of the innovation, emerging when users perceive complications related to the use of the technology (Lee and Kim, 2022). In context of FRAs, perceived barriers emerge when investors view these AI-based technologies as overly opaque, unreliable, or complex due to their algorithmic foundations. This characterization of FRAs as black box lacking transparency in decision-making process leads to a situation of unpredictability. This perception might further provoke cognitive discomfort, confusion, or anxiety to use a new innovation. Previous literature also supports this image barrier results in users’ resistance to use mobile banking (Laukkanen and Kiviniemi, 2010), service robots (Lee and Kim, 2022), autonomous delivery vehicle (Lyu et al., 2024), and IOT (Mani and Chouk, 2018). Perceived image barriers in robo-advisory services emerge when users view these automated platforms as overly complex, opaque, or unreliable due to their algorithmic foundations. The characterization of FRAs as “black boxes” lacking transparency in their decision-making processes contributes to a perception of unpredictability and risk. This might lead to perceiving the AI-driven financial tools as inconsistent, impersonal in comparison to the traditional methods of financial advice. Notably, these image-related doubts about AI often build up specifically when individual beliefs undermine the utility of AI-driven financial services and do not align with their self-concepts and values, which may pose as a psychological reluctance toward these services. Yet, if an individual having a positive disposition toward AI might override the symbolic incongruities linked to FRAs. The more the strong positive attitude and belief toward the AI technology, the more will be the probability to offset the image-related barriers and enhance the openness to use technological alternatives such as FRAs. Thus, attitude of an individuals could play an important role in influencing the degree to which the image barriers can be translated toward resistance to use FRAs. Building on the above discussions, the following hypotheses are formed:

Inertia refers to the tendency to stick with the existing system despite of using other better alternatives and showing resistance to change (Zhang et al., 2024). This inertia can impact the individuals, enhancing beyond functional barriers to address psychological obstacles to innovation. The psychological inertia typically has a tendency to maintain status quo (Samadi et al., 2024). In other words, inertia stems from the unique challenge to reframe the existing ideas and established traditions on the basis of innovative ideas (Schmid, 2019). In context of FRAs, inertia can actually reduce the tendency to believe in the algorithmic intelligent systems leading to psychological changes and favor their existing systems (Danneels et al., 2018). In other words, the investors might stick to their traditional advisory and self-directed methods which might create performance doubts about the FRAs. Previous studies have also validated this, such as healthcare professional resistance (Zhang et al., 2024), AI chatbots (Xi, 2024), and socio-technical inertia (Schmid, 2019). Still, research suggests that this impact might not be same for all the individual investors. There is a possibility that the investors with a positive outlook toward AI are more open to re-establish the routines and may perceive AI-driven systems as an empowering tool rather than disruption (Pal et al., 2025). Their belief system in AI-based technologies such as FRAs may help them counter their inertia that is rooted in traditional practices. Thus, attitude toward the AI might moderate the relationship between inertia and resistance to use FRAs either by reinforcing or dampening their intentions to use FRAs. Building on the same, the following hypotheses can be formed:

Resistance denotes the psychological state of reluctance or aversion caused by conflicting ideas or beliefs when confronted with novel systems, which has a significant influence on user behavior in technology adoption contexts (Mani and Chouk, 2022). This reluctance may stem from skepticism about the financial decision-making by the algorithms or may be due to discomfort with reduced human interactions or may be due to perceived threats. This dissonance might cause negative emotions such as anxiety or confusion, which can further impact their behavioral intentions. Such dissonance not only leads to resistance but also reduces an investor’s intention to accept, further lowering down their willingness to even recommend these services to other users. Previous studies have shown resistance negatively impacts on the user willingness to continue mobile apps (Migliore et al., 2022). Similarly, a study on technology renewal revealed that reasons for user resistance and new information technology (Shirish and Batuekueno, 2021), resistance to change in healthcare (Shahbaz et al., 2019), and another study on O2O platforms resistance by the small retailers (Chawla et al., 2024a; Chawla et al., 2024b; Jafri et al., 2025) have positive impact on discontinuous intentions. Prior attitudes toward AI-backed technologies are a significant predictor of user behavior, influencing how individuals process new information and whether they accept or reject it (Li, 2023). When users have positive prior attitudes towards AI-backed technologies, they are more likely to regard these platforms as efficient and trustworthy, thereby motivating for greater engagement. On the other hand, users with unfavorable or less prior attitudes may feel reluctant due to their conflicting beliefs about the transparency of FRAs. This emotional stress can lead to lowering their willingness to accept these AI-backed financial services. In context of FRAs, users might face cognitive dissonance when they feel uncertainty about the robo-advisors’ platforms to deliver relevant and reliable information conflicting with the desire to stick to their traditional practices. Building on this, we propose the following hypotheses:

In the world of finance and technology services, where investors frequently feel anxious and uncertain about their finances, robo advisory services which are entirely based on algorithms driven advice system may provoke resistance.

Form the above laid literature, the current studies formulate this conceptual model (see Figure 1).

To test the conceptual framework, the study uses India as a geographical setting. Measurement scales from previously published research were used to operationalize the scale items of the identified components. Using the earlier research of Parissi et al. (2019), items of perceived complexity were assessed. The scale items for inertia and perceived security risk were assessed from Mani and Chouk (2018). The standardized measures for over-confidence were taken from Meyer et al. (2013). Items from the research by Chawla et al. (2024a) and Chawla et al. (2024b) were used to get the scale items for the construct value barrier and image barrier. The resistance to use scale items was obtained from research conducted by Chawla et al. (2024a) and Mani and Chouk (2018). The Sindermann et al. (2021) and Cheng et al. (2019) study served as the source of the standardized measures for attitude toward AI. Prior research by Chawla et al. (2024a) was used to evaluate the items of use intention. The items for recommendation intention were extracted from the study of Rahi et al. (2018). A five-point Likert scale was employed to measure each statement, with 1 denoting “strongly disagree” and 5 denoting “strongly agree.”

To assess the items chosen for the investigation, a screening test was conducted with an expert panel consisting of three professors from a reputed state university, as subject-matter experts and four Fintech industry experts from North Indian states. To arrive at the final instrument, two to three discussion sessions were held both in person and electronically via Google Meet. Industry experts recommended the inclusion of individual’s attitude toward AI. In the second step, pilot testing was undertaken, with the questionnaire distributed to a total of 70 research scholars and academics via Google Form as well as in person. This cohort was asked to score the scale items and recommend any things that may be added or alter to increase clarity. A total of 54 respondents filled out the form and offered a few changes to the phrasing of scale items. There were no new scale elements introduced, but four were eliminated due to low mean values. The group’s comments were integrated into the questionnaire to increase its clarity.

Data collection was conducted both online and offline. Non-probability sampling approaches were used due to the lack of a sufficient sampling frame (Vehovar et al., 2016). Data were gathered using purposive and snowball sampling methods. The study opted for the non-probability purposive sampling as the study focused on the investors that were familiar with financial robo-advisors. Still, this method does not provide the full generalization, but the study considered to include the diverse group of respondents and applied very clear screening criteria. In addition, the study also acknowledged the possible biases and carefully interpreted the results, following the widely accepted guidelines (Hair et al., 2010). Purposive sampling is preferable in studies when respondents have little familiarity with the event under investigation (Sibona and Walczak, 2012). Because obtaining a list of all persons using or considering utilizing a financial robo-advisor was tough, data were obtained using the purposive and snowball sampling techniques proposed by Naderifar et al. (2017). The questionnaire was administered via platform such as LinkedIn, WhatsApp, and Telegram, particularly in forums and groups related to financial investment. Nonetheless, we sought individuals through personal and professional networks who had used or preferred to use FRA. Participants were also invited to share the survey link with others with similar profiles or recommend a few colleagues, relatives, and acquaintances, allowing for snowball sampling. Several reminders through emails and revisits were undertaken to approach the respondents. The sample size was chosen using Siddiqui et al. (2013) study, which concluded that 384 respondents are sufficient for up to 90 scale items. Accordingly, approximately 500 plus questionnaires were sent, and after confiscating unfinished responses, a final sample of 409 responses was obtained, with the respondents’ profiles presented in Table 1. This sample size exceeds the allowed limit of 384 and supports the robustness of the statistical analyses and findings. To exclude social desirability biases, the respondents were assured secrecy for their responses. They were also assured that their replies would only be utilized for scholarly reasons. The study’s goals were explained to participants at the outset, and a detailed explanation of robo-advisors was provided to encourage true and candid responses. In addition, respondents were given assurances of data security, anonymity, and confidentiality to protect their privacy and foster confidence. Table 1 depicts the demographic characteristics of the research participants. The bulk of responses are male (71.51%). In terms of age distribution, approximately 60% are between the ages of 25 and 44. Approximately half of respondents (48.17%) had at least a bachelor’s degree, with 34.47% holding a master’s degree. The profile also includes statistics for approximately 48% of respondents with a bachelor’s degree, followed by 34.47 as master degree holders. Furthermore, the majority (33.25%) claimed 1–3 years of investing experience, followed by 27.38% with 4–6 years of experience.

Various prerequisite quality checks were performed before simply moving on to final data analysis. These tests were followed by the measurement model and structural equation modeling was performed using SPSS and AMOS. As recommended by Byrne (2013), the arithmetic mean was used to replace the missing data. Data were checked for non-response bias and to do that the mean differences between the initial 50 responses and the final 50 responses out of all were assessed. Non-response bias was not an issue as no statistical difference was found. The data were further tested to see whether common method bias (CMB) exists or not. Harman’s single-factor test was utilized to test this (Harman, 1976; Shkoler and Tziner, 2017). Assuming that the occurrence of either a general factor or a single factor accounting for the majority of covariance among measures indicates the existence of CMB. This entails: combining all scale items into a single factor with varimax rotation in exploratory factor analysis (Podsakoff et al., 2003, p. 889). It is recommended that the single-factor solution’s explained variance must not exceed 50% (Harman, 1976). The outcomes shown in Appendix Table A.1 display a single factor variance value of 28.570%, which is less than the suggested value. This indicates the absence of CMB.

To assess the collinearity among constructs, the study additionally evaluated for multicollinearity in accordance with the recommendations proposed by Hair et al. (2021). The variance inflation factor (VIF) values were computed to do this. The range of VIF values required to exhibit multicollinearity is 0.20 to 5.0. All of the VIF values, as shown in Appendix Table A.2, fall between 1.29 and 2.38, which is the optimal range and indicates that multi-collinearity is not a problem.

Confirmatory factor analysis (CFA), which indicates hypothesized causal connections between latent and observed indicator variables, was used to make sure fit among observed data and a theoretically grounded model to ensure reliability and validity criteria prior to the structural model assessment (Hancock and Mueller, 2001, p. 5240). The factor loadings of the constructs and average variance extracted (AVE) were taken into consideration to assess the convergent validity of the exogenous and endogenous constructs (Hair et al., 2010). Items with standardized factor loadings of 0.6 or above are considered appropriate (Kline, 2005). Adequate convergent validity is indicated by AVE values more than 0.5 (Bagozzi et al., 1991; Fornell and Larcker, 1981). Composite reliability was calculated to address internal consistency, i.e., reliability. According to Fornell and Larcker (1981), the value of 0.7 for composite reliability is appropriate. In addition, as recommended by Hair et al. (2010), correlation was carried out to verify discriminant validity. According to Hair et al. (2010), the suggested values of AVE should be greater than inter-item correlations, indicating that constructs are not heavily associated. Following the determination of a suitable factor structure, structural equation modeling was used to evaluate the hypothesized correlations between exogenous and endogenous constructs in the study.

From Table 2, it is evident that item loadings for all constructs range from 0.703 to 0.939, which are over the recommended threshold value, i.e., 0.60 (Kline, 2005). All of the scale items’ critical ratio values are greater than 1.96, suggesting that the data are normally distributed (Byrne, 2013). Convergent validity is thus demonstrated by these findings. In addition, Table 2 reports on the validity and reliability of the constructs and the scale items associated with them. The composite reliabilities of the constructs, which fall between 0.842 and 0.951, all exceed the threshold of 0.7, illustrating internal consistency. The fact that each construct’s AVE is higher than 0.5 confirmed convergent validity and further supports overall model validity (Fornell and Larcker, 1981).

To ensure discriminant validity, correlation analysis of the constructs was performed. To ensure the discriminant validity, the square root of AVE must be more than inter-item correlations, which means the constructs are not highly correlated (Hair et al., 2010). Table 3 demonstrates that the square root of all AVE values is higher than the inter-item correlations and hence indicates that the measurement model has sufficient validity and the model is appropriate for further structural testing.

To test the hypotheses, moderation test was performed. Table 5 indicates the effects of moderation where attitude toward AI does not moderate the relationship of perceived complexity, value barrier, and image barrier with resistance. This indicates that user’s having favorable attitude toward AI is not immune to challenges such as complex interfaces or stereotype associated with FRAs. Significant moderating effects of attitude toward AI were seen in the cases of inertia, overconfidence bias, and data privacy risk. This implies that users’ resistance is less affected by the data privacy risk if they have supportive disposition to AI. Because of this, users’ perceived vulnerability may be lessened (Oprea et al., 2024). In addition, in case of overconfidence bias, user’s reluctance toward FRAs is reduced as their favorable attitude toward AI will strengthen their belief in the superiority of AI-backed decision-making (Heidari, 2024). Finally, the relationship between inertia and resistance is also moderated by attitude toward AI, inferring that positive AI beliefs could alter habitual resistance and drive openness to change (Balakrishnan et al., 2024).

On the psychological basis, user resistance was prominently influenced by overconfidence bias, image barrier, and inertia. These findings revealed that cognitive bias, identity associations, and a dependence on conventional financial practices continue to impede widespread adoption of robo-advisors’ platforms. Essentially, resistance itself had a negative and significant influence on both the usage intentions and recommendation to FRAs, validating resistance as a key barrier in user decision-making process. The study presents a nuanced and empirically validated knowledge of the key barriers driving reluctance to financial robo-advisor adoption in Indian context. The findings, drawn from a sample of 409 respondents and analyzed using structural equation modeling (SEM), confirm that both functional and psychological blockades influence resistance to robo-advisory services. Among the functional barriers, data privacy risk and perceived complexity were shown to considerably increase user reluctance, underscoring ongoing concerns about usability and the management of sensitive financial data. Conversely, value barrier which is often seen as a central factor of innovation resistance was found to be minimal. This finding demonstrates rising consumer awareness and knowledge of the low-cost, accessibility, and other concrete benefits of robo-advisory platforms, specifically when contrasted to conventional human financial advisors.

In addition, to enhance the model, the study tested the moderating influence of attitude toward AI on the relation between individual barriers and resistance. Noteworthy, this moderating impact was only observed in cases of data privacy risk, overconfidence bias, and inertia, implying that those individuals with a more positive attitude toward AI are less resistant, even in the face of these specific problems. Nevertheless, no significant moderation was discovered for value barrier, perceived complexity, and image barrier, suggesting that technological optimism alone may not be enough to overcome perceived usability difficulties or social-symbolic reluctance.