Why do consumers avoid AI clones? The role of perceived inauthenticity, ideological incompatibility and negative consumer–influencer relationships

E-commerce companies are turning to artificial intelligence (AI) bots as a cost-effective alternative to human hosts, leveraging their ability to live stream continuously without fatigue or time limitations (Yiying and Ziqing, 2024). The current AI industry in live streaming is rapidly expanding, especially in China, where virtual influencers and non-playable character style creators are becoming popular across platforms like Douyin (TikTok) (Gastel and Ryder, 2023). The virtual human industry is anticipated to experience remarkable growth, with projections estimating its value to reach $440.3bn by 2031 (Allied Market Research, 2023). Alongside this growth, the rise of AI influencers represents a pivotal development in marketing strategies, where brands are increasingly adopting AI-driven tools to improve their efficiency and reach. In the realm of human–AI communication, it is crucial to acknowledge that AI systems have evolved beyond being simple tools; they now function as communicators and collaborative partners across diverse interactions (Sarwari et al., 2024).

Specifically, within the domain of AI integration into live-streaming e-commerce, organisations encounter substantial risks when AI influencers are improperly used. A major issue is the lack of authenticity that virtual influencers present, which can alienate consumers. A specific example is the AI live streaming of Calvin Chen, where despite the label indicating AI usage, he lost 7,000 followers within days (Gastel, 2024), highlighting the negative impact of perceived inauthenticity on audience engagement. While there has been extensive research on positive consumer–brand relationships (Aw et al., 2022), the “dark side” of these relationships, particularly negative feelings and behaviours, has not been as thoroughly explored (Brandão and Popoli, 2022). Compared to traditional influencers, AI influencers provide distinct benefits, such as cost-effectiveness and reduced risks of reputation-related controversies, positioning them as a reliable option for brands aiming for stability and control in their campaigns (Sands et al., 2022). Nevertheless, despite these potential benefits, limited research exists on how consumers perceive and interact with AI influencers. While preliminary findings indicate that audiences may respond favourably to AI-generated content, deeper insights into their trust levels and emotional connections with such influencers remain largely unexplored (Sands et al., 2022).

In addition, the emergence of virtual influencers and AI clone influencers is different from real human influencers, raising ethical questions during the interaction between consumers and influencers (Robinson, 2020). Brands need to be aware of the ethical implications of their virtual influencers’ behaviours (Mouritzen et al., 2024). Consumers may perceive AI influencers as “authentically fake”, which can challenge their desire for authenticity in influencer interactions (Lou et al., 2023). In addition, a mismatch of the value between the consumer and brand will produce negative emotions for consumers (Rodrigues et al., 2021). Therefore, does the AI clone influencer contradict the consumer’s values or mismatch the consumer’s moral considerations? A study indicates that the perceived inauthenticity of AI influencers leads to consumers perceiving the brand as insincere or disconnected from their values and expectations (Alboqami, 2023). However, there exists a research gap in AI clone avoidance.

AI clone avoidance refers to the deliberate behaviour of consumers who choose to evade or disengage from AI clone live-streaming platforms (Hegner et al., 2017b). Consumers often exhibit a negative bias towards AI, which can lead to distrust and avoidance behaviours when interacting with these technologies (Jobin et al., 2019). Research indicates that social anxiety can moderate individuals’ willingness to accept AI technologies, as consumers may feel uncomfortable interacting with systems that mimic human behaviour (Yuan et al., 2022). Ethical considerations are a critical factor in AI clone avoidance, as consumers are becoming increasingly cognisant of the ethical challenges associated with AI technologies. These challenges include concerns about privacy, informed consent and the potential for AI systems to reinforce existing biases (Jobin et al., 2019).

This research aims to investigate consumer avoidance behaviour by negative consumer–influencer relationships due to ideological incompatibility on the perceived inauthenticity of AI clone influencers. This research considers a moral perspective to investigate factors facilitating AI clone avoidance. The results can help identify the ethical dilemmas that consumers face when interacting with AI clones, allowing developers to create guidelines and frameworks that prioritise ethical considerations in AI design and implementation. In addition, the result of this research will help manager in luxury industry to have a better understanding of consumer behaviour towards AI clone influencer.

This study advances existing knowledge beyond the dominant assumption that “inauthenticity is uniformly detrimental” by demonstrating that different dimensions of perceived inauthenticity do not exert equivalent effects on consumer responses. Specifically, while prior research has largely treated inauthenticity as a unidimensional construct leading to negative outcomes, this study reveals a more nuanced mechanism in which only integrity and reliability violations trigger ideological incompatibility, whereas continuity, credibility and symbolism do not. This differentiation challenges the prevailing “inauthenticity is bad” narrative and introduces a more granular understanding of how consumers morally evaluate AI-driven agents. In doing so, the study contributes to emerging research on AI influencers by identifying which aspects of inauthenticity are ethically consequential and which are merely perceptual, thereby refining both authenticity theory and AI ethics discourse.

The following structure of this paper will first introduce this study’s theoretical background: the moral competency theory and the conception of the negative consumer–influencer relationship. After that, the development of the hypothesis will be presented. Next, the research methods and results will be presented. Following that, the paper will discuss the theoretical and practical contributions of the study. Finally, the limitations of the study and recommendations for future research will be discussed.

The moral competency theory indicates that consumers’ behaviour is influenced by their sense of justice (Kohlberg, 1976; Phau et al., 2009). This theory suggests that consumers are not merely driven by self-interest but are also motivated by moral imperatives that reflect their beliefs about fairness and justice in the marketplace. The moral competency theory also emphasises that consumers use moral competencies, such as ethical reasoning and integrity, to make decisions. Their actions are guided by personal beliefs about what is right and wrong, and this influences their consumption choices. For example, consumers may choose ethical products or avoid companies with unethical practices, based on their internal sense of justice and morality (Morales-Sánchez and Cabello-Medina, 2013). Previous research has shown that when a brand’s values or actions do not align with a consumer’s personal beliefs, morals or societal expectations, it can lead to negative emotions such as brand hate or consumer boycotts. Consumers may see the brand’s position on legal, ethical or social issues as conflicting with their own, leading them to reject the brand, even if they previously supported it (Ren et al., 2024). This research uses the moral competency theory to explain consumers’ negative relationships with AI clone influencers.

Recent developments in AI ethics, including frameworks on algorithmic transparency and explainability (e.g. DARPA’s explainable AI initiative), further reinforce the importance of accountability in AI-driven interactions.

According to the theory of moral responsibility, individuals are regarded as moral agents if they can be held accountable for their actions, provided those actions are performed voluntarily and with an understanding of their potential consequences (Fischer, 1986). This theory underscores the importance of interpersonal relationships, as society expresses reactive attitudes such as blame and praise based on individuals’ actions within these relationships (Eshleman, 2014).

In the context of influencer marketing, both influencers and brands are regarded as moral agents, bearing the responsibility to adhere to and uphold ethical standards. When they fail to do so, particularly in the eyes of Generation Z, they can be perceived as morally irresponsible, leading to negative consumer reactions such as brand avoidance (Pradhan et al., 2023). This study aims to examine the effects of perceived inauthenticity in AI clone influencers and the influence of ideological incompatibility. The theory of moral responsibility serves as a foundation for developing hypotheses regarding how ideological incompatibility impacts consumers’ negative relationships with AI clone influencers and their avoidance behaviours towards AI clone influencers.

The concept of perceived inauthenticity originates from the notion of brand authenticity, which reflects the extent to which consumers view a brand as genuine and aligned with its core values (Bruhn et al., 2012). Perceived inauthenticity refers to the subjective experience in which individuals or groups feel that their expressions, actions or identities do not align with their true selves or internal beliefs (Silver et al., 2021). Perceived inauthenticity of AI clone influencers describes the diminished authenticity consumers feel when they know content or interactions come from AI clones instead of genuine human input. The study divided the perceived inauthenticity of AI clone influencers into five distinct dimensions: continuity, credibility, integrity, symbolism and reliability. This framework is adapted from established models of brand authenticity, notably those proposed by Bruhn et al. (2012) and Morhart et al. (2015), which identify key facets such as continuity, originality, reliability, credibility, integrity and symbolism.

Continuity refers to the extent to which consumers perceive the stability, longevity and consistency of an AI clone’s presence and behaviour (Bruhn et al., 2012). Credibility pertains to the degree to which consumers believe in the authenticity and truthfulness of the AI clone (Morhart et al., 2015). Integrity reflects the extent to which consumers view AI clones as guided by principles of care and responsibility (Morhart et al., 2015). Symbolism represents the ability of AI clones to enable consumers to express and remain true to their own identities (Morhart et al., 2015). Reliability refers to the degree to which consumers perceive the trustworthiness, credibility and keeping promises of the AI clone (Bruhn et al., 2012).

A low level of brand authenticity indicates that the brand promise does not originate from its core values (Schallehn et al., 2014). As a result, consumers are likely to view brand actions as inauthentic, because the brand promise has not been fulfilled in a personalised, continuous and consistent way (Rodrigues et al., 2021; Schallehn et al., 2014). Past research also indicates that perceptions of the inauthenticity of a brand have prompted consumers to avoid the brand (Thompson and Arsel, 2004). Non-authentic brands may enhance consumers’ negative emotions towards a brand (Rodrigues et al., 2021; Thompson and Arsel, 2004). In addition, stance continuously evolves along the brand–authenticity continuum is an important factor of brand authenticity (Napoli et al., 2016). Lee et al. (2009) noted that when a brand becomes overly mainstream, loses its exclusivity or fails to retain consumer respect, it may be perceived as generic or inauthentic. In addition, individuals often react negatively to perceived inauthenticity, viewing it as a moral violation (Silver et al., 2021). According to the moral competency theory, moral justice plays an important role in consumer behaviour (Kohlberg, 1976; Phau et al., 2009). Thus, the perceived inauthenticity will enhance consumers’ feelings of moral violation. To have a better understanding of the relationship between each dimension of perceived inauthenticity and consumers’ ideological incompatibility, the following hypotheses are presented:

Ideological incompatibility refers to a collection of beliefs that conflict with the consumer’s own values (Hegner et al., 2017b). This incompatibility is a key factor influencing brand hate, alongside negative past experiences and symbolic misalignment (Hegner et al., 2017a). When consumers recognise a disconnect between their personal values and those conveyed by an influencer or brand, it can lead to the intention to boycott (Wang et al., 2021). Another research focus on how political ideology significantly influences consumer choices, suggesting that consumers’ preferences can be swayed by the ideological leanings of the products or influencers they engage with (Ho, 2022). Research has shown that ideological incompatibility significantly impacts symbolic incongruence (Otoo et al., 2023). Another study highlights that both ideological and symbolic incongruence can contribute to brand hate (Islam et al., 2019). Similarly, Mohammed et al. (2022) found that ideological incompatibility significantly influences celebrity brand hate, which, in turn, drives negative word-of-mouth. In addition, according to the theory of moral responsibility, if the influencer violates consumer moral standards, the consumer will behave negatively (Fischer, 1986; Pradhan et al., 2023). Based on the previous discussion, the following hypotheses are proposed:

A negative consumer–influencer relationship refers to a detrimental dynamic between influencers and consumers (Reinikainen et al., 2021; Shao et al., 2024). This research identifies the negative consumer–influencer relationship with AI clones, including non-engagement, negative word-of-mouth and hate emotion.

This perspective aligns with foundational work on the dark side of consumer–brand relationships, which highlights how relational breakdowns can trigger retaliation, avoidance and negative emotional responses (Fournier and Alvarez, 2012; ,Grégoire and Fisher, 2008). Furthermore, this dynamic can be interpreted through the lens of parasocial relationships, where consumers form one-sided emotional bonds with influencers (Horton and Wohl, 1956). AI clone influencers may disrupt these relationships by introducing perceived artificiality, thereby weakening emotional attachment and intensifying negative responses.

Consumer engagement refers to the active participation of customers in their interactions with a brand or firm, leading to enhanced psychological states of self-identification with the brand (Blasco-Arcas et al., 2016). Non-engagement refers to the unfavourable feelings, thoughts and behaviours consumers exhibit towards a brand, often resulting from negative experiences (Rodrigues et al., 2021). This concept contrasts with positive influencer engagement, which involves cognitive, emotional and behavioural investment in AI clone interactions.

In this study, negative word-of-mouth refers to the extent to which individuals engage in negative discussions or criticisms of AI clones (Hegner et al., 2017b; Rodrigues et al., 2021). In addition, it is defined as any informal communication between private parties that expresses unfavourable opinions or evaluations of AI clones (Rodrigues et al., 2021). It serves as a warning mechanism for other consumers, aiming to dissuade them from supporting a particular AI clone provider. Hate emotion refers to a range of aversive emotional states with AI clones (Dabbous and Aoun Barakat, 2023).

Hate emotion refers to a range of aversive emotional states with AI clones (Dabbous and Aoun Barakat, 2023). Hate is characterised as a complex emotional response, often seen as a more intense form of dislike. Hate can encompass various feelings, including anger, disgust and disappointment, which may trigger both passive and active responses towards the object of hate. Negative consumer–brand interactions often result in consumer subversion, wherein consumers actively resist, reject or distance themselves from the brand (Kuanr et al., 2022). Based on the previous discussion, the following hypotheses are presented:

This research focuses on consumer responses to AI clone live streaming, a trend that is rapidly gaining traction in China. Popular platforms such as Taobao Live, Douyin (the Chinese version of TikTok) and JD Live are frequently used for AI-powered live streaming (Yiying and Ziqing, 2024). Consequently, the sample for this study consists of live-streaming consumers in China, with the target population being Chinese individuals who engage with live-streaming platforms. A screen question was designed to ensure target population inclusion: “Have you watched or interacted with any live-streaming content (e.g. shopping, gaming, entertainment) on Chinese platforms (e.g. Taobao Live, Douyin, Bilibili) at least once in the past month?” Respondents who answer “No” and those whose addresses are not in China will be excluded from the survey. The questionnaire was distributed through WJX (Link to wjxLink to the cited website of wjx), a popular questionnaire platform appropriate for accessing diverse samples in China (Del Ponte et al., 2024). To give respondents a better understanding of AI clone live streaming and the news of AI clones of Calvin Chen’s live streaming (Gastel, 2024), the questionnaire begins with an introduction to AI clone live streaming. Thus, the respondents will clearly understand the AI clone and complete the questionnaire based on their direct response to the news of Calvin Chen’s AI clone live streaming.

There are three parts of the questionnaire, the first part is the introduction of AI Clone live streaming, and the example of Calvin Chen’s AI clone live streaming. The second part is the demographic questions. The third part is the measurement items of all variables. The seven-point Likert scale is used for designing questionnaires, participants were asked to rate their responses on a scale ranging from 1 (Strongly Disagree) to 7 (Strongly Agree) for each statement.

All the measurement items are adapted or adopted based on past studies. The variables under the perceived inauthenticity of AI clone influencers (Continuity, Credibility, Integrity, Symbolism and Reliability) are reverse-coding based on brand authenticity (Bruhn et al., 2012; Morhart et al., 2015 and adapt the measurement items from the consumer perceptions of a band into the consumer perceptions of AI clone influencers. Ideological incompatibility has four items that were adapted from Hegner et al. (2017). Non-engagement was reverse-coded and adapted based on the original item of customer engagement from Blasco-Arcas et al. (2016). Negative word-of-mouth, which has three items, was adapted from Wen-Hai et al. (2019). The original item of Hate emotion was Brand hate, which has six items (Hegner et al., 2017b). AI clone live-streaming avoidance has four items, and all are adapted based on the original items of brand avoidance (Hegner et al., 2017b). The details of all measurement scales are shown in Table 1. A pilot test of the questionnaire was conducted with 60 participants. For all variables, Cronbach’s α values were higher than 0.74, which suggested acceptable internal consistency (Hair et al., 2019).

The minimum sample size was calculated by G power, with nine predictors (CO, CR, IN, SY, RE, IN, NE, NW, HE and AVO), f² = 0.15, α = 0.05 and efficacy = 0.95, the minimum sample size in this research is 166 (Faul et al., 2009). This research has collected 329 effective responses, which exceeded the requirements of minimum sample size. As shown in Table 2, a significant majority of the respondents are females (83.3%) compared to males (16.7%). Most are young, with 92.1% being under 21 years old. Educational attainment is predominantly at the junior college level (93.0%), while a small portion holds bachelor’s degrees (2.4%) or master’s degrees (1.8%). In terms of income, 91.2% earn less than ¥5,000 per month, while only 2.4% earn more than ¥20,000. The frequency of live-streaming consumption shows that 54.7% watch less than once per week, while 14.6% watch more than four times per week.

The partial least squares structural equation modelling (PLS-SEM) approach was employed using SmartPLS software version 4.0 to evaluate the study’s measurement and structural models. Mardia’s multivariate skewness (β = 34.54) and kurtosis (β = 213.94) both yielded p-values below 0.001, confirming that the data deviated from a multivariate normal distribution. As a result, PLS-SEM was deemed appropriate for the data analysis in this research (Yan et al., 2021).

Common method bias (CMB) often arises when survey respondents provide responses for multiple variables within a single survey (Kock et al., 2021). To assess CMB, this study applied the common method factor analysis recommended by Liang et al. (2007). As shown in Table 3, the ratio of the average Ra² to the average Rb² was sufficiently high (75.351), suggesting that common method bias is negligible.

Table 4 summarises the constructs used in this study, along with their factor loadings, Dijkstra-Henseler’s rho (rho-a), composite reliability (CR) and average variance extracted (AVE). Factor loadings indicate the strength of the correlation between each item and its respective construct, with most values exceeding the recommended threshold of 0.7 (Hair et al., 2017). Although CO1 has a factor loading of 0.549, the AVE for the CO construct is 0.563, which exceeds the 0.5 threshold, confirming acceptable convergent validity (Byrne, 2016). CR values for all constructs are above 0.7, demonstrating strong internal consistency (Hair et al., 2018). In addition, all AVE values are greater than 0.5, further supporting adequate convergent validity (Fornell and Larcker, 1981). In addition, discriminant validity was assessed using the heterotrait–monotrait (HTMT) ratio. As indicated in Table 5, all confidence interval values are below 1, establishing discriminant validity in this study (Hair et al., 2018).

Table 6 presents variance inflation factor (VIF) values for different constructs in a structural model. Typically, a VIF value above 10 indicates potential issues (Vittinghoff et al., 2005). VIF values below 10, indicate that multicollinearity is less of a concern for these paths. Afterwards, a bootstrap procedure with 5,000 iterations was used to evaluate the hypothesis. As shown in Table 7, H1c (β = 0.142, p < 0.05), H1e (β = 0.676, p < 0.001), H2a (β = 0.142, p < 0.05), H2b (β = 0.377, p < 0.001), H2c (β = 0.661, p < 0.001), H3a (β = 0.276, p < 0.001) and H3c (β = 0.707, p < 0.001) are supported. On the contrary, H1a, H1b, H1d and H3c are unsupported in this research.

The f-square values presented in Table 8 illustrate the effect sizes of various predictor variables on the dependent constructs. In SEM, f-square quantifies the extent to which a predictor variable influences an endogenous variable. According to Cohen (1988), f-square values are categorised as follows: 0.02 indicates a small effect, 0.15 indicates a medium effect and 0.35 indicates a large effect. Firstly, continuity (f² = 0.008), credibility (f² = 0.005), integrity (f² = 0.016) and symbolism (f² = 0.014) have no effect on ideological incompatibility. Reliability (f² = 0.357) has a large effect on ideological incompatibility. Secondly, ideological incompatibility has a large effect on hate emotion (f² = 0.775), a medium effect on negative word-of-mouth (f² = 0.166) and a large effect on non-engagement (f² = 5.546). Finally, hate emotion (f² = 0.602) has a large effect, negative word-of-mouth (f² = 0.012) has no effect and non-engagement (f² = 0.166) has a medium effect on AI clone avoidance.

Table 9 presents Q-square and R-square values for different variables, offering insights into their predictive relevance and explained variance. The Q-square values are used to evaluate the predictive relevance of the model, with a Q-square value greater than 0 indicating significant predictive ability (Hair et al., 2017). As shown in Table 9, all Q-square values exceed 0, demonstrating the strong predictive relevance of the model’s paths. Conversely, R-square values represent the proportion of variance explained by the model. An R-square value closer to 1 indicates a strong relationship between variables, whereas a value closer to 0 suggests a weaker relationship (Zikmund, 2010). AI clone avoidance has an R-square of 0.724, explaining 72.4% of the variance. Hate emotion has a relatively low R-square of 0.437, showing weaker explanatory power. Ideological incompatibility leads to an R-square of 0.827, indicating the strongest explanatory power among the variables. Negative word-of-mouth has the lowest R-square at 0.142, showing low explanatory power. Finally, non-engagement has an R-square of 0.847, reflecting high explanatory power.

Firstly, there are two dimensions of perceived inauthenticity of AI clones that are supported to have a positive impact on Ideological Incompatibility: Integrity (H1c) and Reliability (H1e). Perceived lack of integrity in the AI clone influencers fosters ideological incompatibility as consumers may feel that the AI clone influencers do not align with their values or beliefs, leading to a rejection of the AI clone influencers’ outputs or recommendations. Research indicates that when individuals perceive a lack of integrity in a brand or entity, they are more likely to develop negative attitudes towards it, which can manifest as brand hate or avoidance (Bryson and Atwal, 2019; Rodrigues et al., 2021; Sharma et al., 2023). Furthermore, lack of reliability, which encompasses the inconsistency or unpredictability of AI clones, can similarly exacerbate feelings of ideological incompatibility. Consumers may find it challenging to reconcile their beliefs with the outputs of an unreliable AI, leading to cognitive dissonance that fuels negative perceptions (Bayarassou et al., 2020). When integrity and reliability are compromised, consumers are more likely to feel that the AI does not represent their ideological stance, prompting a rejection of its use.

On the contrary, the other dimensions, Continuity (H1a), Credibility (H1b) and Symbolism (H1d) do not have a significant impact on ideological incompatibility.

Lack of continuity may not directly challenge consumers’ ideological beliefs as much as the integrity and reliability dimensions do. Similarly, perceived lack of credibility and symbolism may not invoke the same level of ideological conflict because they do not directly challenge the user’s core beliefs or values. Ideological incompatibility often arises from a perceived misalignment between a brand’s values and those of the consumer, leading to avoidance behaviours (Hegner et al., 2017b).

Secondly, ideological incompatibility has a significantly positive impact on the negative consumer–influencer relationships: H2a, H2b and H2c are all supported.

When influencers or brands are perceived as misaligned with a consumer’s ideology, it can lead to adverse reactions, including brand hate and avoidance. Furthermore, ideological incompatibility is shown to foster negative sentiments towards influencers who fail to resonate with consumers’ core beliefs (Brandão and Popoli, 2022). Thus, ideological incompatibility plays a critical role in shaping negative consumer–influencer relationships. When consumers detect a misalignment between their personal beliefs and the values promoted by influencers or brands, it can result in substantial negative outcomes such as Non-engagement, negative word-of-mouth and the development of hateful emotions.

Finally, non-engagement (H3a) and hate emotion (H3c) both have a positive impact on AI clone avoidance; on the contrary, negative word-of-mouth (H3b) does not have a significant impact on AI clone avoidance. When users perceive AI clones as inauthentic or unreliable, they may choose to disengage entirely, leading to avoidance behaviours. Research indicates that disengagement is often a protective mechanism used by consumers to shield themselves from negative experiences or emotions associated with a brand or product (Tsarenko and Tojib, 2012). Hate emotion, on the other hand, represents a more intense emotional response that can drive consumers to actively reject or avoid AI clones. This aligns with the concept of brand hate, where consumers develop a strong aversion to a brand due to perceived betrayals or failures, leading to avoidance and negative behaviours towards the brand (Zarantonello et al., 2016). In contrast, negative word-of-mouth does not significantly impact AI clone avoidance. This may be attributed to the nature of negative word-of-mouth itself, which often serves as a form of communication rather than a direct behavioural response. While negative word-of-mouth can influence perceptions and attitudes towards a brand, it does not necessarily compel individuals to avoid the brand outright. For instance, consumers may engage in negative word-of-mouth as a way to express dissatisfaction or seek validation from peers without necessarily withdrawing from the brand or product.

Firstly, this research contributes to the theory of moral competency by highlighting the role of moral judgments in consumer behaviour through the examination of perceived inauthenticity and its relationship with ideological incompatibility. And illustrates how perceptions of authenticity influence moral evaluations and subsequent actions, reinforcing the notion that moral competency is not merely about the ability to judge but also about the capacity to act in alignment with those judgments (Silver et al., 2021). Furthermore, it emphasises the need for brands and influencers to cultivate authenticity in their interactions with consumers, as failure to do so can lead to significant moral backlash and disengagement (Burgess and Jones, 2023).

Secondly, this research advances the theory of moral responsibility by demonstrating how perceived inauthenticity in AI clones can affect consumer behaviour and relationships. It emphasises the significance of ethical considerations in the development and application of AI, as well as the moral duties of influencers to uphold authenticity in their interactions with consumers. In addition, it highlights how ideological beliefs influence consumer perceptions and actions, thereby reinforcing the need for a deeper understanding of the complex relationship between technology, morality and consumer behaviour (Hegner et al., 2017b).

Firstly, the insights gained from understanding perceived inauthenticity can guide the design and development of AI systems. Companies can prioritise the creation of AI clones that embody the values and beliefs of their target audiences, thereby reducing ideological incompatibility. This approach can involve incorporating user feedback into the development process, ensuring that AI systems are designed to resonate with the cultural and ideological contexts of their users. By doing so, organisations can enhance the perceived authenticity of their AI clones, leading to improved user satisfaction and loyalty.

On the other hand, to counteract the negative effects of ideological incompatibility, companies in the AI clone industry should prioritise authenticity in their influencer strategies. This involves selecting influencers who genuinely align with the brand’s values and who can communicate authentically with their audience. By fostering genuine relationships between influencers and consumers, companies can reduce the perceived inauthenticity of the AI clone influencers, thereby reducing the likelihood of avoidance behaviours.

Beyond managerial implications, this research also offers important societal and policy-related insights. The findings highlight the ethical risks associated with AI clone influencers, particularly when issues of integrity and reliability are not adequately addressed. Regulators and policymakers may use these insights to inform the development of ethical guidelines for AI-driven marketing, including clearer disclosure requirements for AI-generated content, standards for transparency in algorithmic communication and accountability mechanisms for AI-enabled interactions. Such measures can help mitigate consumer distrust and promote more responsible use of AI technologies in digital marketplaces.

A key limitation of this study lies in the demographic composition of the sample, which is heavily skewed towards young, female and junior college–educated respondents. While this segment represents an important and active group within live-streaming consumption, it does not fully reflect the broader population of Chinese consumers. Prior research suggests that age, gender and education can significantly influence technology adoption, trust in AI and ethical evaluations. As such, the observed relationships in this study may differ across demographic groups, limiting the generalisability of the findings. Future research should therefore replicate this model using more demographically diverse samples to assess the robustness of the proposed relationships and explore potential moderating effects of age, gender and education.

To overcome these limitations, future research should prioritise a more geographically diverse and robust sampling approach. In addition, future research is suggested to focus on the ethical implications of AI technology. As AI continues to evolve, concerns regarding privacy, data security and algorithmic bias may shape consumer perceptions and behaviours towards AI clones. Given the rise of virtual influencers (Williams, 2024), it is crucial to explore the ethical implications of these technologies on user interactions and emotional responses.