生成AIが作成した広告キャラクターが消費者の広告評価に及ぼす影響

Abstract

In recent years, virtual influencers have emerged. Asia’s first virtual influencer, imma, is active on social media, and appears in advertisements for the luxury brand Coach. Previous studies have noted that virtual influencers are compatible with utilitarian products and that human influencers are preferable for hedonic products. However, there are cases in which virtual influencers are used in advertisements for hedonic products; thus, a gap exists between theory and reality. Therefore, this study conducted an exploratory investigation of consumers’ anthropomorphic tendencies and product knowledge as conditions for virtual influencers to function effectively with hedonic products. The results showed that virtual influencers were effective for consumers with high anthropomorphic tendencies and high product knowledge for certain hedonic products. This study provides new insights into the conventional finding that virtual influencers and hedonic products have low compatibility and offers practical suggestions for the effective use of virtual influencers in advertising hedonic products.

Translated Abstract

近年，バーチャルインフルエンサーが登場している。アジア初のバーチャルインフルエンサーであるimmaは，SNSで活躍し，さらには，ラグジュアリーブランドCoachの広告に起用された。先行研究では，バーチャルインフルエンサーは実用財との適合度が高く，快楽財においては人間のインフルエンサーのほうが好ましいと指摘されてきた。しかし，現実には，バーチャルインフルエンサーが快楽財の広告に使用される事例も見られ，理論と実態に乖離が存在している。そこで，本研究では，快楽財において，バーチャルインフルエンサーが効果的に機能する条件として，消費者の擬人化傾向と製品知識に着目し，探索的調査を行った。その結果，特定の快楽財において，擬人化傾向が高く，製品知識が高い消費者においては，バーチャルインフルエンサーの有効性が示された。本研究は，バーチャルインフルエンサーと快楽財は適合度が低いとする従来の知見に対して新たな知見を提供するとともに，快楽財におけるバーチャルインフルエンサーの効果的活用に向けた実務的示唆を提示した。

I.  Introduction

Many companies use influencers to promote their products and brands on social networking sites. An influencer is a person who has expertise in a particular field and influences others’ purchasing decisions (De Veirman et al., 2017; Lou & Yuan, 2019). In the past, influencers were predominantly real people. However, in recent years, virtual influencers have emerged. Virtual influencers are computer-generated fictional persons with human characteristics, traits, and personalities (Thomas & Fowler, 2021).

In fact, imma, Asia’s first virtual influencer, is active on social media platforms such as X, Instagram, and TikTok. She not only runs her own accounts but also appears in advertisements for the luxury brand Coach. Prior research has indicated that consumers perceive AI, such as virtual influencers and chatbots, as logical and more connected to practical values, although they find it difficult to connect with AI influencers emotionally (e.g., Huang & Rust, 2018). Zhang et al. (2022) showed that in the case of utilitarian products, AI-designed products increase consumers’ willingness to pay (WTP), whereas in the case of hedonic products, human-designed products increase consumers’ WTP. In other words, it is more effective to use virtual influencers when the advertised product is a utilitarian product and human influencers when the advertised product is a hedonic product.

However, the findings of previous studies that virtual influencers are less compatible with hedonic products do not always apply (Belanche et al, 2024). In fact, in the case of imma, virtual influencers have been used in advertisements for hedonic products, such as fashion goods, and have attracted a certain amount of attention, which contrasts with the findings of previous studies. Accordingly, the level of compatibility between virtual influencers and hedonic products may be suppressed by certain factors. Therefore, the objective of the study is to identify the conditions under which virtual influencers are as effective as or more effective than human influencers in promoting hedonic products. The research question is:

RQ: Are there conditions under which virtual influencers are as effective in advertising hedonic products as human influencers are?

To answer this research question, we focus on two consumer characteristics: anthropomorphic tendencies and product knowledge. Anthropomorphic tendencies refer to individual characteristics that tend to anthropomorphize nonhuman objects (Letheren et al., 2016; Waytz et al., 2010). This study measures the tendency to anthropomorphize AI. Consumers with a greater tendency to anthropomorphize AI are more likely to view virtual influencers as human-like (Biason et al., 2024) and perceive less incongruity with hedonic products. When consumers with greater product knowledge focus on central product information cues when evaluating advertisements, they give less attention to the influencer’s appearance (Gammoh et al., 2018). Therefore, their evaluations are less influenced by the influencer’s appearance, thus discrepancies are less likely to be perceived.

The aim of this study is to identify the consumer characteristics that enable virtual influencers to have the same or greater effect than human influencers in the case of hedonic products. It addresses previous research limitations and contributes to theoretical development. Identifying consumer segments compatible with virtual influencers for hedonic products is expected to have practical applications.

II.  Literature Review

In recent years, virtual influencers have been active on social media platforms such as Instagram and have gained global attention as enablers of interactive and personalized communication between brands and consumers (Audrezet & Koles, 2023). Previous studies have shown that the attractiveness of virtual influencers positively impacts brand attachment (Kim & Park, 2023), consumer product engagement and purchase intention (PI) (Biason et al., 2024). Furthermore, virtual influencers are of interest to companies as a cost-effective marketing method because they do not age, are easy to manage visually, and are not subject to scandals or other risks that often plague the use of human influencers and celebrities. However, virtual influencers still have limitations compared with human influencers. Human influencers have the advantage of building relationships through emotional empathy. Hence, this study reviews previous studies comparing human and virtual influencers and aims to clarify the differences in their influence on advertising.

1.  Virtual vs. human influencers

Recent studies have increasingly compared virtual and human influencers in advertising. Virtual influencers are often viewed as less effective because of their limited sensory ability and lower credibility, resulting in weaker influencer attitudes (Wan & Jiang, 2023) and brand attitudes (BAs) (Ozdemir et al., 2023).

While some studies claim that virtual influencers are less effective than human influencers are, others claim that virtual influencers are more effective. For example, virtual influencers were found to be more effective than humans in recommending experience products (Zheng et al., 2024) or utilitarian products (Belanche et al., 2024), indicating that interaction with product type is an important factor. In particular, usefulness is the main motivating factor for virtual influencers, whereas identification is the main motivating factor for humans (Belanche et al., 2024).

In summary, although virtual influencers are generally considered less effective in advertising than human influencers are, their effectiveness may be somewhat moderated by factors such as autonomy, uniqueness, aesthetic imperfection, and usefulness (Song et al., 2024; Volles et al., 2024). The interaction with product type (search vs. experience, utilitarian vs. hedonic) is also an important factor in understanding the advertising effects of virtual influencers.

2.  Consumer characteristics: Anthropomorphic tendencies

Anthropomorphic tendencies are an important element for understanding the advertising effectiveness of virtual influencers. Anthropomorphic tendencies are a psychological characteristic of individual consumers and refer to the psychological tendency to anthropomorphize a subject. In other words, it is the tendency to naturally apply the cognitive bias of anthropomorphism (Letheren et al., 2016; Waytz et al., 2010).

Many previous studies have examined the effects of anthropomorphizing objects such as products (Aggarwal & McGill, 2007), brands (Vernuccio et al., 2025), tourist attractions (Lin et al., 2024), and robots (Yang et al., 2024) on consumer behavior. However, the degree to which an object is anthropomorphized as an individual characteristic of consumers has not been sufficiently examined. The former refers to giving a specific object a “shape”, such as a human-like appearance or characteristics, whereas the latter refers to the psychological mechanism of consumers that occurs to the object. Letheren et al. (2016) noted a gap in this research and described the importance of understanding the individual characteristics of consumers’ anthropomorphic tendencies, in addition to the causes, processes, and consequences of anthropomorphism.

In this study, we focus on anthropomorphic tendencies as a factor that makes people perceive humanness in virtual influencers. According to social identity theory, people perceive those who belong to the same category as themselves as members of an inner group. They evaluate the inner group more positively than they do the outer group (Tajfel & Turner, 1986). Thus, consumers with high anthropomorphic tendencies will perceive virtual influencers created by generative AI as similar to human influencers and will be less likely to feel uncomfortable with products or brands recommended by these influencers. Previous studies have shown that individuals with high anthropomorphic tendencies respond more positively to anthropomorphic cues in advertisements (Letheren et al., 2016). Additionally, Li and Sung’s (2021) study of attitudes toward smart speaker-type AI assistants revealed that consumers with greater anthropomorphic tendencies evaluated assistants more positively.

The stronger an individual consumer’s anthropomorphic tendencies are, the more likely they are to accept virtual influencers as human beings, perceive them as trustworthy, and evaluate products recommended by virtual influencers more favorably.

3.  Product knowledge

Product knowledge refers to consumers’ general knowledge of a product category (Mason et al., 2001). Product knowledge is used by consumers when purchasing and using products and is considered to play an important role in decision-making.

According to the elaboration likelihood model (ELM), consumers who are knowledgeable about a product are less influenced by influencers with peripheral information about the product (Petty et al., 1983). This relationship has also been confirmed for nonhuman entities.

Gammoh et al. (2018) demonstrated that when product knowledge is poor, consumers give excessive attention to the appearance of the avatar, strongly perceiving a discrepancy with the product category and reducing PI. In contrast, consumers with good product knowledge focus on the main product information and are less influenced by the avatar’s appearance, which mitigates the effect of the discrepancy. On the basis of the above discussion, we can infer that consumers’ product knowledge affects their perceptions of the discrepancy between virtual influencers and hedonic products.

III.  Survey and Analysis

1.  Research framework and scales

This study aimed to identify the conditions under which virtual influencers are as effective or more effective than human influencers in the case of hedonic products. Therefore, this study uses analysis of variance to examine how stimuli from human or virtual influencers affect consumers’ advertising evaluations and then, using the mean values of the dependent variables obtained from the analysis, identifies the conditions under which virtual influencers are as effective or more effective than human influencers in the case of hedonic products. The independent variables used in this study were the type of influencer (human/virtual), while the moderating variables were anthropomorphic tendencies and product knowledge. Age, income, and curiosity were used as control variables. Curiosity (Olney et al., 1991) was considered a control variable because it influences the evaluation of virtual influencers and hedonic products (Zhang et al., 2022).

We considered four aspects as dependent variables: influencer attractiveness (IA), attitudes toward the advertisement (AA), BA, and PI. The dependent variables are defined as follows: IA indicates the extent to which consumers perceive human or virtual influencers as attractive (Von Mettenheim & Wiedmann, 2021). AA refers to the degree to which consumers like or are interested in an advertisement (e.g., Biehal et al., 1992). BA represents consumers’ favorable or negative evaluations of a particular brand (Kumar et al., 2024). PI refers to the intention to purchase a product or brand in the future (Kumar et al., 2024). Studies 1 and 2 were conducted to target hedonic products. To confirm robustness, the study used experimental products: branded cosmetics (foundation) in Study 1 and branded bags in Study 2.

Considering the possibility that the use of AI technology and virtual influencers may be met with resistance among some generations, the subjects of this experiment were individuals in their 20s and 30s who were relatively familiar with these technologies. Because purchasing luxury cosmetics and fashion products may be uncommon among males, participation was limited to females.

2.  Study 1

(1)  Experimental design

In Study 1, cosmetics were employed as hedonic products, and a fictitious advertisement was used for the experiment. This study employed a 2 (type of influencer: human/virtual)×2 (anthropomorphic tendency: high/low)×2 (product knowledge: high/low) between-subjects experimental design. Two types of advertisements were created for the experiment, featuring a human influencer and a virtual influencer. The number of participants was 1120. The participants were asked to answer a series of questions to confirm that they had fully read the instructions and recognized the type of influencer. Specifically, they were asked questions such as “The image you saw was a virtual influencer created by AI (yes, no);” “The image you saw was of a living human influencer (yes, no);” and only data from participants who answered these questions correctly were analyzed (N=487). At the end of the response, we confirmed that the images viewed were fictitious products and brands and that the images of virtual influencers were processed by generative AI to avoid misunderstanding among the respondents.

(2)  Experimental materials and procedures

In this experiment, we employed an unknown female as the human influencer. The virtual influencers were created using Adobe Firefly. Specifically, photos of the human influencers used in the experiment were input into the application and processed and edited to obtain the appearance of a virtual human. Cosmetics were selected as the experimental products. The virtual brand name “LUNORIA” was used for the experimental product to eliminate the influence of brand recognition on the experimental participants. The advertisements for the experiment included the product brand name, product package image, and influencer. Appendix A presents the advertisements used in the experiments.

For the experiment, two types of magazine advertisements were created: (type of influencers: human/virtual). The participants were randomly assigned to one of the two patterns. As part of the experimental procedure, a questionnaire titled “Survey on Influencers and Product Advertising” was first presented to the participants (Appendix B).

(3)  Results

In Study 1, confirmatory factor analysis (CFA) was used to test the construct validity of the variables. The measurement items with factor loadings below 0.7 (anthropomorphic tendency: “AI has an agenda;” product knowledge: “I use cosmetics frequently;” and IA: “This influencer has a neat look”) were excluded to increase the accuracy of the measurement model. The final measurement model showed a good fit (χ²(255)=695.983, p<.001, CMIN/DF=2.73, GFI=0.89, CFI=0.96, TLI=0.96, IFI=0.96, RMSEA=0.06). To evaluate convergent and discriminant validity, this study calculates the average variance extracted (AVE) and composite reliability (CR), following the methods of Fornell and Larcker (1981). The results confirmed that each scale was sufficiently reliable (Appendix B).

Multivariate analysis of covariance (MANCOVA) was conducted to examine the overall impact of the independent variables on the dependent variables (IA, AA, BA, and PI). The results revealed that influencer type (Pillai’s Trace=.07, F(4, 473)=8.42, p=.000, partial η²=.066), anthropomorphic tendency (Pillai’s Trace=.04, F(4, 473)=4.49, p=.001, partial η²=.037), and product knowledge (Pillai’s Trace=.13, F(4, 473)=17.51, p=.000, partial η²=.129) had significant multivariate effects. The interaction terms and second-order interaction terms were either marginally significant or not significant. The mean values listed below indicate that, although there are exceptions, being human, having a high degree of anthropomorphism, and having high product knowledge generally increased the values of the dependent variables (Table 1). The sample size varied from 36 to 88 participants; thus, Box’s M test was significant (p=.000). However, each cell had over 30 participants, so the MANCOVA was robust to cell size imbalance and deviations from equal variance (Tabachnick & Fidell, 2013).

Table 1

Differences in means between human and virtual influencers in Study 1

Note: Covariates are ratings fixed at the following values. Age=30.42; Household income=6.02; Curiosity=2.90. Adjustment for multiple comparisons: Bonferroni. *=p<.05.

Next, analyses of covariance (ANCOVAs) were performed to confirm the effects on each dependent variable. Unlike with the other dependent variables, however, the main effect of influencer type on PI was not significant (p=.206). This suggests that influencers become less important at the purchase stage. Appendix C presents the main results.

The results indicate that among consumers with high anthropomorphic tendencies and product knowledge, there are no significant differences in any of the dependent variables between human and virtual influencers (Figure 1d). Additionally, among consumers with high anthropomorphic tendencies and low product knowledge, no differences were observed between human and virtual influencers in any dependent variable except influencer attractiveness (Figure 1c). These findings suggest that differences between human and virtual influencers are less likely to emerge among consumers with high anthropomorphic tendencies.

Figure 1

Comparison of human and virtual influencers by conditions (cosmetics)

3.  Study 2

(1)  Experimental design

In Study 2, bags were employed as hedonic products, and an experiment was conducted using fictitious advertisements. The experimental design and types of influencers were the same as those used in Study 1. The number of participants in the experiment was 1120. After viewing the experimental advertisements, the participants were asked to answer the same questions as in Study 1. The analysis was based exclusively on data from participants who correctly identified the influencer type (N=566).

(2)  Experimental materials and procedures

The method for creating virtual influencers and the experimental procedure are the same as those in Study 1. The fictitious brand name used for the product in the experiment was “CEVALOIRE.” The questions were also modified to fit the context of bags and used the same scale as in Study 1 (Appendix B).

(3)  Results

As in Study 1, CFA was conducted in Study 2 to verify the construct validity of the variables used. The results of the analysis revealed that the measurement model demonstrated a good fit (χ²(255)=695.406, p<.001; CMIN/DF=2.73; GFI=0.91; CFI=0.97; TLI=0.96; IFI=0.97; and RMSEA=.06). Additionally, as shown in Appendix B, the reliability of each scale was confirmed.

The MANCOVA revealed significant multivariate effects on influencer type and product knowledge (Pillai’s Trace=.12, F(4, 552)=18.17, p=.000, partial η²=.116; Pillai’s Trace=.06, F(4, 552)=8.81, p=.000, partial η²=.060). As in Study 1, the interaction and second-order interaction terms were either marginally significant or not significant. Table 2 shows means by condition for dependent variables and the differences between human and virtual influencers. The sample size ranged from 35 to 109 participants, with more than 30 participants in each cell.

Table 2

Differences in means between human and virtual influencers in Study 2

Note: Covariates are ratings fixed at the following values. Age=30.36; Household income=5.79; Curiosity=2.80. Adjustment for multiple comparisons: Bonferroni. *=p<.05.

Unlike Study 1, the ANCOVAs revealed that the influencer type (F(1, 555)=4.40, p=.036, partial η²=.008) also had a significant effect on PI (Appendix C). Unlike in Study 1, the results of Study 2 showed that human influencers were more effective when consumers had high levels of anthropomorphism and product knowledge (Figure 2d). Conversely, regardless of product knowledge, there was no difference between human and virtual influencers in terms of BA and PI when the anthropomorphic tendencies were low (Figure 2a and 2b).

Figure 2

Comparison of human and virtual influencers by condition (bags)

IV.  Discussion

This study explored the effects of influencer type (human/virtual), consumer anthropomorphic tendency, and product knowledge on advertising effectiveness measures (IA, AA, BA, and PI) for hedonic products. In both Studies 1 and 2, the results of the multivariate analysis revealed that being a human influencer and having high product knowledge were significant factors. Additionally, in terms of PI, it was confirmed that there were many conditions in which no difference was found in the dependent variable values between human and virtual influencers. These results are relatively robust for hedonic products. This study investigated the conditions identified in Study 1, which examined cosmetics as hedonic products. The study revealed that consumers with high anthropomorphic tendencies and extensive product knowledge did not perceive significant differences between human and virtual influencers. These results suggest that anthropomorphism-prone consumers may be affected by virtual influencers in the same way as they are by human influencers. This aligns with social identity theory. Previous studies have shown that anthropomorphism-prone consumers perceive virtual influencers as equivalent to humans (Letheren et al., 2016). According to the ELM, consumers with considerable product knowledge show high product involvement and base their evaluations on central cues, such as information about the product itself, rather than peripheral cues, such as the influencer’s appearance (Gammoh et al., 2018).

In contrast, Study 2 focused on the bags, which are also hedonic products. In contrast to Study 1, when anthropomorphic tendencies and product knowledge were high, human influencers presented higher values than did virtual influencers in all the dependent variables. These results can be explained by the difference between public and private products within the category of hedonic products. Because others seldom see hedonic products for private use, the motive to value the credibility and social reputation of the recommending influencer is weaker, and the effect of the influencer type on PI is less. In contrast, for public hedonic products, human influence significantly affects AA, suggesting that credibility and social evaluation are critical when people are conscious of the evaluation made by others.

Furthermore, in the case of public hedonic products, there was no difference between human and virtual influencers in terms of BA and PI under conditions with low anthropomorphic tendencies. Virtual influencers are regarded as machines or tools, separate from the social evaluation of brands. This suggests that anthropomorphic tendencies may have opposite effects on private and public products, even among hedonic products. These results differ from those shown by Li and Sung (2021) and presented a new aspect of anthropomorphic tendencies.

V.  Conclusion

1.  Theoretical contributions

This study makes two theoretical contributions. First, it identifies conditions under which virtual influencers can produce effects comparable to those of human influencers in advertising hedonic products. While previous studies have recognized the attractiveness, uniqueness, and usefulness of virtual influencers (Biason et al., 2024; Volles et al., 2024), their suitability for hedonic products remains an issue (Zhang et al., 2022). This study expanded research on virtual influencers by verifying their effectiveness in marketing hedonic products.

Second, based on social identity theory and the ELM, this study identified consumer characteristics associated with the perception of virtual influencer effectiveness: anthropomorphic tendencies and product knowledge. In response to the criticism by Letheren et al. (2016) about the lack of understanding of individual characteristics, this study deepened the understanding of psychological mechanisms in virtual influencer research. This study provides a new perspective that will contribute to future theory building.

For private products, consumers with high anthropomorphic tendencies and product knowledge perceive virtual influencers as human-like, which mitigates the discrepancy between virtual influencers and hedonic products. This is the first study to obtain this result.

Conversely, for public products, since the evaluation by others is a critical factor, even consumers with high anthropomorphic tendencies tend not to evaluate advertisements that feature virtual influencers favorably. This suggests that a more detailed classification of conditions is necessary to determine whether virtual influencers have the same effect as human influencers on hedonic products. Thus, this study focused on product attributes, which have not been adequately considered in previous studies, and their influence on the effects of virtual influencers. In this context, this study contributes new findings to the literature.

2.  Practical implications

Our results have the following practical implications. First, firms should consider consumers’ anthropomorphic tendencies when selecting influencers. Consumers with high anthropomorphic tendencies give favorable ratings to advertisements and brands featuring virtual influencers, whereas human influencers are more effective in advertising private products for consumers with low anthropomorphic tendencies. For example, the active use of virtual influencers would be effective when the target consumers are those with a high affinity for character-like entities. This is because a high affinity for character-like entities indicates a greater tendency to anthropomorphize nonhuman objects. Second, firms should consider the product attributes and context of product use when selecting influencers.

3.  Limitations and future challenges

Despite the contributions of this study, it has several limitations. First, since the sample was limited to women in their 20s and 30s, the results may not be generalizable to other ages, genders, or lifestyles. Therefore, studies of more diverse consumer groups are needed. Second, we employed an unnamed human influencer and an original virtual counterpart to control for appearance. Their anonymity might have shaped their perceptions of IA and attitudes. Future studies should use well-known influencers to test how influencer awareness affects evaluations. Third, virtual influencers are computer-generated fictional characters with human characteristics, traits, and personalities. The stimuli used in this study followed this procedure; however, the similarity of the stimuli to illustrations is a limitation. Fourth, this work was exploratory. This study revealed that anthropomorphism and product knowledge moderate responses to virtual influencers, but it did not formally test theory-driven hypotheses. Subsequent research should strengthen the theoretical rigor by developing and testing hypotheses based on persuasion models. Addressing these issues and building on these findings will deepen our understanding of virtual influencer advertising effectiveness.

Acknowledgments

This paper was written in Japanese, roughly translated into English using DeepL, and carefully edited by the author. We would like to thank Editage (www.editage.jp) for English language editing. The experimental stimuli were processed with Adobe Firefly, a generative AI program trained on copyright-free images.

Data Availability

The data for this study are available from the J-STAGE Data repository.

• 1) Effects of influencer type on customer evaluations (Study 1: Cosmetics)

• 2) Effects of influencer type on customer evaluations (Study 2: Bags)

References

•  Aggarwal,  P., &  McGill,  A. L. (2007). Is that car smiling at me? Schema congruity as a basis for evaluating anthropomorphised products. Journal of Consumer Research, 34(4), 468–479. https://doi.org/10.1086/518544
• Audrezet, A., & Koles, B. (2023). Virtual influencer as a brand avatar in interactive marketing. In C. L. Wang (Ed.). The Palgrave handbook of interactive marketing (pp. 353–376). Springer International Publishing. https://doi.org/10.1007/978-3-031-14961-0_16
•  Belanche,  D.,  Casaló,  L.v., &  Flavián,  M. (2024). Human versus virtual influences, a comparative study. Journal of Business Research, 173, 114493. https://doi.org/10.1016/j.jbusres.2023.114493
• Biason, R., Elnagar, A. K., Tolete, C., Elsaadany, H. A. S., & others. (2024). Virtual influencer marketing and purchase intention: The mediating roles of product involvement and brand familiarity [Preprint]. Research Square. https://doi.org/10.21203/rs.3.rs-5324811/v1
•  Biehal,  G.,  Stephens,  D., &  Curio,  E. (1992). Attitude toward the ad and brand choice. Journal of Advertising, 21(3), 19–36. https://doi.org/10.1080/00913367.1992.10673373
•  De Veirman,  M.,  Cauberghe,  V., &  Hudders,  L. (2017). Marketing through Instagram influencers: The impact of number of followers and product divergence on brand attitude. International Journal of Advertising, 36(5), 798–828. https://doi.org/10.1080/02650487.2017.1348035
•  Fornell,  C., &  Larcker,  D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.2307/3151312
•  Gammoh,  B. S.,  Jiménez,  F. R., &  Wergin,  R. (2018). Consumer attitudes toward human-like avatars in advertisements: The effect of category knowledge and imagery. International Journal of Electronic Commerce, 22(3), 325–348. https://doi.org/10.1080/10864415.2018.1462939
•  Huang,  M.-H., &  Rust,  R. T. (2018). Artificial intelligence in service. Journal of Service Research, 21(2), 155–172. https://doi.org/10.1177/1094670517752459
•  Kim,  H., &  Park,  M. (2023). Virtual influencers’ attractiveness effect on purchase intention: A moderated mediation model of the Product–Endorser fit with the brand. Computer in Human Behavior, 143, 107703. https://doi.org/10.1016/j.chb.2023.107703
•  Kumar,  S.,  Rajaguru,  R., &  Yang,  L. (2024). Investigating how brand image and attitude mediate consumer susceptibility to eWOM and purchase intention: Comparing enterprise-owned vs. third-party online review websites using multigroup analysis. Journal of Retailing and Consumer Services, 81, (Article 104051). https://doi.org/10.1016/j.jretconser.2024.104051
•  Laroche,  M.,  Bergeron,  J., &  Goutaland,  C. (2003). How intangibility affects perceived risk: The moderating role of knowledge and involvement. Journal of Services Marketing, 17(2), 122–140. https://doi.org/10.1108/08876040310467907
•  Letheren,  K.,  Kuhn,  K.-A. L.,  Lings,  I., &  Pope,  N. K. L. (2016). Individual difference factors related to anthropomorphic tendency. European Journal of Marketing, 50(5/6), 973–1002. https://doi.org/10.1108/EJM-05-2014-0291
•  Li,  X., &  Sung,  Y. (2021). Anthropomorphism brings us closer: The mediating role of psychological distance in User–AI assistant interactions. Computers in Human Behavior, 118, 106680. https://doi.org/10.1016/j.chb.2021.106680
•  Lin,  C. J. Z.,  Ruan,  H.,  Zhang,  H., &  Xiao,  S. X. (2024). Antecedents, manifestations, and environmental consequences of destination anthropomorphism: The case of nature-based destinations. Journal of Hospitality and Tourism Management, 58, 432–444. https://doi.org/10.1016/j.jhtm.2024.03.001
•  Lou,  C., &  Yuan,  S. (2019). Influencer marketing: How message value and credibility affect consumer trust of branded content on social media. Journal of Interactive Advertising, 19(1), 58–73. https://doi.org/10.1080/15252019.2018.1533501
•  Mason,  K.,  Jensen,  T.,  Burton,  S., &  Roach,  D. (2001). The accuracy of brand and attribute judgments: The role of information relevancy, product experience, and attribute-relationship schemata. Journal of the Academy of Marketing Science, 29(3), 307–317. https://doi.org/10.1007/BF02890787
•  Olney,  T. J.,  Holbrook,  M. B., &  Batra,  R. (1991). Consumer responses to advertising: The effects of ad content, emotions, and attitude toward the ad on viewing time. Journal of Consumer Research, 17(4), 440–453. https://doi.org/10.1086/208569
•  Ozdemir,  O.,  Kolfal,  B.,  Messinger,  P. R., &  Rizvi,  S. (2023). Human or virtual: How influencer type shapes brand attitudes. Computers in Human Behavior, 145, 107771. https://doi.org/10.1016/j.chb.2023.107771
•  Petty,  R. E.,  Cacioppo,  J. T., &  Schumann,  D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 10(2), 135–146. https://doi.org/10.1086/208954
•  Song,  X.,  Lu,  Y., &  Yang,  Q. (2024). The negative effect of virtual endorsers on brand authenticity and potential remedies. Journal of Business Research, 185, 114898. https://doi.org/10.1016/j.jbusres.2024.114898
• Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics (6th ed.). Pearson.
• Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel, & W. G. Austin (Eds.). Psychology of intergroup relations (pp. 7–24). Nelson-Hall.
•  Thomas,  V. L., &  Fowler,  K. (2021). Close encounters of the AI kind: Use of AI influencers as brand endorsers. Journal of Advertising, 50(1), 11–25. https://doi.org/10.1080/00913367.2020.1810595
•  Vernuccio,  M.,  Boccalini,  S., &  Patrizi,  M. (2025). “Being” with the brand in the metaverse: Strengthening brand anthropomorphism to foster brand love. Journal of Retailing and Consumer Services, 84, 104204. https://doi.org/10.1016/j.jretconser.2024.104204
•  Volles,  B. K.,  Park,  J.,  Van Kerckhove,  A., &  Geuens,  M. (2024). How and when do virtual influencers positively affect consumer responses to endorsed brands? Journal of Business Research, 183, (Article 114863). https://doi.org/10.1016/j.jbusres.2024.114863
•  Von Mettenheim,  W., &  Wiedmann,  K. P. (2021). The role of fashion influencers’ attractiveness: A gender-specific perspective. Communication Research and Practice, 7(3), 263–290. https://doi.org/10.1080/22041451.2021.2013087
•  Wan,  A., &  Jiang,  M. (Montina). (2023). Can virtual influencers replace human influencers in live-streaming e-commerce? An exploratory study from practitioners’ and consumers’ perspectives. Journal of Current Issues & Research in Advertising, 44(3), 332–372. https://doi.org/10.1080/10641734.2023.2224416
•  Waytz,  A.,  Cacioppo,  J. T., &  Epley,  N. (2010). Who sees human?: The stability and importance of individual differences in anthropomorphism. Perspectives on Psychological Science, 5(3), 219–232. https://doi.org/10.1177/1745691610369336
•  Yang,  Y.,  Chi,  M.,  Bi,  X., &  Xu,  Y. (2024). How does the anthropomorphism of service robots impact employees’ role service behavior in the workplace? International Journal of Hospitality Management, 122, 103857. https://doi.org/10.1016/j.ijhm.2024.103857
•  Zhang,  H.,  Bai,  X., &  Ma,  Z. (2022). Consumer reactions to AI design: Exploring consumer willingness to pay for AI-designed products. Psychology & Marketing, 39(11), 2171–2183. https://doi.org/10.1002/mar.21721
•  Zheng,  X.,  Miao,  M.,  Zhang,  L., &  Huang,  L. (2024). Standing out or fitting in? How perceived autonomy affects virtual influencer marketing outcomes. Journal of Business Research, 185, 114917. https://doi.org/10.1016/j.jbusres.2024.114917

Appendices

Appendix A

(Study 1 Cosmetics Human/Virtual/Study 2 Bags Human/Virtual)

Appendix B

(Study 1 Cosmetics/Study 2 Bags)

Appendix C

(Analysis of Covariance for Study 1 Cosmetics/Study 2 Bags)

Note: For multivariate tests, only the most robust Pillai’s Trace is listed; for ANCOVA, Bonferroni’s adjustment α=.05 was applied. Effects with p≥.05, except the main one, were omitted. Interpretation criteria for partial η²: .01=small, .06=medium, .14=large. In Study 2, Levene’s test was significant (p=.002) for one variable (PI); however, the effect of equivariance was deemed minor due to the small difference in the maximum number of cells.

Hikaru Makino

Hikaru Makino is an Associate Professor in the Faculty of Business Administration at Kindai University. He holds a Ph.D. in Business Administration from Ritsumeikan University. Before his current role, he was an assistant professor, lecturer, and associate professor in the Faculty of Economics, Kanazawa Seiryo University. His primary research interests are behavioral experience and embodiment.

Yuumi Ogura

Yuumi Ogura is an assistant professor at Chukyo University. Her main research areas are marketing, advertising, and consumer behavior. She holds a Ph.D. in Business Administration from Ritsumeikan University Graduate School of Business Administration. She has been in the current position since 2025.

Xiuyan Yan

Xiuyan Yan (Ph.D., Ritsumeikan University) is an Assistant Professor at Hiroshima Shudo University, Japan. Her research interests are in the areas of international marketing, consumer behavior and brand management. Her research has appeared in the International Journal of Emerging Markets, the Journal of Retailing and Consumer Services, and others.