Consumer Evaluation of Virtual vs. Human Influencers via Source Credibility, Perceived Social Similarity, and Consumption Motivation

Abstract

Influencer marketing is estimated to reach USD 32.55 billion by the year 2025. The extant research on virtual vs. human influencers’ effectiveness has reported mixed results. Currently, research has yet to explore how consumption motivation and influencer gender (virtual vs. human) may differentially impact consumer behavior. Likewise, whether antecedent factors such as prior attitudes toward the brand may affect the perceived authenticity and attractiveness of influencers has rarely been investigated. To fill this research gap, the current study examined virtual vs. human influencers’ effectiveness utilizing a 2 (influencer type) × 2 (product type) × 2 (influencer gender) between-subject online experiment. Specifically, Airbnb (a recreational product) and NJM (an insurance product) were respectively designated as a hedonic and utilitarian brand. The findings (N = 468) demonstrated that while prior brand attitudes toward a hedonic product had no effect on perceived influencer authenticity, the opposite was true for a utilitarian product. No significant difference was shown in perceived authenticity and attractiveness between a male and female (virtual or human) influencer. Structural equation modeling suggested that perceived social similarity between a participant and an influencer positively impacted the perceived attractiveness and authenticity of influencers and purchase intention. Hedonic instead of utilitarian motivation was found to be a positive predictor of purchase intention.

1. Introduction

Social media influencers can play a significant role in facilitating marketing communication to affect consumers’ brand attitudes and purchase intentions [1]. Influencer marketing is expected to reach approximately USD 32.55 billion by the year 2025 [2]. For Gen Ziers, following social media influencers is a way of “socializing with friends”, as it evolves around the alignment of hobbies, personal values, and lifestyles between followers and their influencers—also known as Key Opinion Leaders [3].

Past research on the effectiveness of virtual influencers reported mixed results, as virtual influencers were found to either outperform [4,5] or underperform [6] relative to their human counterparts. For instance, prior work has indicated that consumers perceived virtual influencers to be less authentic than human influencers, which negatively impacted their attitude toward the brand and their purchase intentions [7]. However, additional studies found that virtual influencers could attract consumer attention and foster a brand’s visibility and recognition [5,8], in addition to being equally as effective as their human counterparts in affecting consumer intentions to adopt product recommendations [9,10,11].

As the product type can help determine purchase decision-making [12,13], only less than a handful of studies examined the effectiveness of either a virtual or human influencer as an endorser for more than one type of product [7,8,9]. For instance, Belanche et al. [9] revealed that while consumers considered virtual influencers to be more useful in endorsing utilitarian than hedonic products, they identified with human influencers at a “human” level. Other scholars (e.g., [7]) also found that consumers perceived virtual influencers as being less authentic than human influencers across all product types (functional vs. symbolic vs. experiential).

Even though utilitarian vs. hedonic products were evaluated in past influencer studies, research has yet to explore the role of utilitarian vs. hedonic motivation in influencer-driven purchase intention [7,9]. While consumer–influencer identification can help shape influencers’ effectiveness, a theory-based explanation of such an identification remains limited [14,15]. Similarly lacking is the empirical research that compares whether male and female influencers (virtual and human) may have a differential effect on consumer behavior [6,8,16]. Extant research also needs to further validate the preliminary work that established the role of antecedent factors—such as product familiarity and prior attitude toward the brand—in affecting consumer perception of influencers’ attributes and their purchase intentions [17].

In essence, to better understand how the growing use of virtual influencers may reshape the social media marketing landscape, additional research needs to address these contradictory findings and the empirical gap identified above. Against this backdrop, the present work conducted an experiment to test a conceptual framework via influencer type (human vs. virtual), influencer gender (male vs. female), and product type (hedonic vs. utilitarian), with Instagram as the social media platform of choice. This conceptual framework explored the role of product familiarity, prior attitude toward product, influencer–follower social similarly, and influencer attributes, as well as hedonic and utilitarian consumption motivation, in consumer decision-making. The study results contribute to theoretical and practical understanding of the effects of virtual vs. human influencers on consumer decision-making.

2. Literature Review

2.1. Influencer Attributes and Purchase Intention

Source credibility can play a vital role in the persuasive effect of a message [18,19,20]. McCroskey and Teven’s [21] source credibility scale was developed to measure source competence, caring/goodwill, and trustworthiness factors in an interpersonal communication context. Ohanian [22] developed a source credibility scale by measuring consumer perception of a celebrity endorser’s expertise, trustworthiness, and physical attractiveness. Applying Ohanian’s widely adopted scale [22], Lin et al. [17] examined a real-life Instafamous endorser and found that both influencer trustworthiness and attractiveness were significant predictors of brand attitude but not influencer expertise. Other studies suggested that an influencer’s physical attractiveness could lead to a higher follower stickiness to the influencer’s posts or videos [20], taste leadership [23], and an influencer–follower parasocial relationship [24].

In the context of social media influencers, perceived influencer authenticity is an important dimension of perceived influencer trustworthiness. Gerlich [25,26] found that trust plays a critical role in influencing virtual influencer’s overall acceptance and ultimate purchase intentions. A systematic review by Byun and Ahn [27] comparing virtual influencers and human influencers noted that social media followers still question virtual influencers’ trustworthiness—due to the perceived lack of influencer autonomy and controllability—in creating the content presented to their followers. In other words, virtual influencers were seen as lacking authenticity, because their creators created their posted content, compared to human influencers who had the ability to express their authentic self. The significance of virtual influencers’ authenticity in marketing success was further evidenced by its ability to predict consumers’ emotional attachment toward and trust in the brand endorsed by the virtual influencer [15].

Conceptually speaking, perceived authenticity reflects whether consumers perceive an influencer as being genuine and true to their beliefs, values, and identity through self-disclosure, spontaneity, and amateurism. By contrast, perceived trustworthiness describes whether consumers consider an influencer as honest, believable, and dependable [28,29]. In essence, authenticity can be a necessary but not a sufficient element of trustworthiness. For instance, prior work revealed that perceived influencer authenticity was a key driver of advertising content’s effectiveness, as it was positively related to perceived influencer trustworthiness and affected followers’ purchase intentions [30]. Other studies also reported that perceived human influencers’ authenticity had a positive effect on followers’ attitudes, willingness to follow, and purchase intentions [7,29].

Based on the literature discussed above, the key attributes of a virtual influencer that are evaluated by social media users are as follows. First, a virtual social media influencer is an unknown “character” to social media users at introduction. Second, everything that appears in the social media profile about the virtual influencer is fictional. Third, when a virtual influencer posts any product endorsements, as well as any comments and responses to followers, the posted content is created by the brand’s marketing team. Fourth, social media users may evaluate the credibility of a virtual influencer via their authenticity and attractiveness more than their trustworthiness and expertise, as they consider interacting with these influencers as a process of “socializing with friends” [31] via parasocial interaction [17].

Even though perceived attractiveness is an important construct in influencer credibility research [17,20], existing literature examining how the physical attractiveness of virtual vs. human influencers may influence their followers’ purchase behaviors is scarce. As previous work on social media influencers has found that influencer authenticity could play a pivotal role in impacting consumers’ purchase behaviors [5,15], it is logical to assume that the perceived attractiveness and authenticity of a social media influencer (virtual or human) could lead to a higher purchase intention.

H1a-b: 

Perceived influencer (a) attractiveness, and (b) authenticity will be positively associated with consumer purchase intention.

2.2. Follower–Influencer Identification and Perceived Social Similarity

Previous studies on influencer communication such as livestreaming, for instance, found that social identity was positively associated with consumers’ purchase behaviors (e.g., [32,33]). Another study [34] also found that consumers’ brand attitudes would be reinforced if the advertised brand was more in congruence with their own self-identity. In particular, consumers are motivated to engage influencers not only because they perceive them as an “in-group” member, but also because of their desire to consider themselves as being similar to the influencers that they admire. This audience behavior is consistent with the assumption of social identify theory [35,36], which asserts that people’s self-concepts are shaped by their social group membership; such a membership (perceived or real) allows them to classify other individuals in society as their in-group—often based on such factors as sociological attributes, religious faith, cultural affinity, and political ideology that inspire them to embrace this shared identity [37].

Perceived similarity is an important factor associated with perceived social identity and perceived social distance between consumers and product endorsers [38]. This construct is particularly significant for the U.S. Gen Z population who have grown up in a diverse multicultural society and transcultural communication environment, where K-pop culture and its stars are accepted and admired as global pop culture icons [39]. As such, research has begun to operationalize perceived similarity beyond its original 20th-century definition to reflect a more contemporary interpretation of social identity via the construct of perceived social similarity. The construct of perceived social similarity emphasizes whether the product spokesperson is similar to someone considered to be in the consumer’s close circle of friends, larger social network, and peer group, in addition to other social identifiers such as geography, race/ethnicity, and nationality [40].

The prior literature has suggested that perceived social similarity between consumers and an advertising spokesperson could affect customers’ trust in the spokesperson [39]. For instance, a study by Xiao et al. [41] demonstrated that perceived similarity between YouTubers and their followers could act as a significant heuristic cue, enhancing the perceived credibility of the YouTubers. In other social media contexts, when influencers share similar interests and language with their followers, they can bridge the gap and foster a sense of intimacy with them [42]. Perceived follower–influencer similarity was also found to increase consumers’ tendency to recommend their favorite influencers to others [43]. Additional research has shown that followers’ perception of their perceived similarity to the influencer could strengthen the parasocial relationship between them [24].

Although the topic of social influencers has been widely studied, the existing research has yet to explore the relationship between perceived influencer similarity and two sources of credibility attributes—influencers’ attractiveness and authenticity. Based on the theoretical and empirical literature reviewed above, it is reasonable to propose the following hypotheses to test these relationships.

H2a-b: 

Perceived social similarity will be positively associated with perceived influencer (a) attractiveness, and (b) authenticity.

2.3. Hedonic vs. Utilitarian Consumption Motivation

The marketing literature [44,45] suggests that both a hedonic (an experiential motivation) and utilitarian consumption motivation (a functional motivation) can help explain consumers’ purchase decision-making process. A hedonic consumption motivation is associated with consumer needs for experiencing emotions, fun, excitement, and pleasure-seeking goals [46]. For instance, consumers could gain instant gratification by making impulse purchases [47,48] or buying products to meet their need for mood management [49]. A utilitarian consumption motivation is linked to consumers’ rational thinking, functional or practical needs, and end-goal considerations [50]. Specifically, utilitarian purchases are driven by meeting consumers’ task-oriented goals, to make their purchase decisions more precisely and efficiently [51]. Compared to hedonic shoppers, utilitarian shoppers obtain more gratification when marketing messages highlight a product’s functional information to help them evaluate those functions and make purchase decisions efficiently [52].

Prior literature has found that both hedonic and utilitarian consumption values play an important role in affecting consumer behaviors in an online shopping environment. For instance, Pahnila and Warsta [53] confirmed that both hedonic and utilitarian values could enhance consumers’ mood and increase their enjoyment of online shopping process, if they had a pleasant shopping experience and transparent product information regarding quality and functionality. Additional research also reported that utilitarian and hedonic motivation could serve as antecedents for continuance intention and purchase intention in the social media context [54,55,56]. Park et al.’s [57] study further indicated that when micro-influencers (those with 10,000 to 100,000 followers) endorsed a hedonic product, they could amplify the relationship between consumer perception of their authenticity and the brand endorsed by them.

As prior research has suggested that consumer behavior is shaped by the interplay between utilitarian and hedonic consumption motivations [44,58], relying on only a single motivation type may overlook the complexity of the consumer’s decision-making process. It is worth noting that a product primarily considered to be hedonic in nature may contain elements that appeal to utilitarian consumption motivation and vice versa [44,58,59]. In the current study, we conceptualize utilitarian and hedonic consumption motivations as two distinct motivational dimensions that primarily and respectively correspond to consumers’ functional and experiential consumption needs [45]. We also anticipate that as a hedonic consumption motivation will have a significant effect on purchase intention for hedonic products, a utilitarian consumption motivation will also have a significant effect on purchase intention for utilitarian products. To validate these conceptual assumptions, we propose the following hypotheses.

H3: 

Hedonic consumption motivation will be positively associated with purchase intentions toward a hedonic product instead of utilitarian consumption motivation.

H4: 

Utilitarian consumption motivation will be positively associated with purchase intentions toward a utilitarian product instead of hedonic consumption motivation.

2.4. Brand Familiarity, Prior Brand Attitude, and Influencer Attributes

The effect of brand familiarity on branding success has long been established in the literature. For instance, prior work has shown that brand familiarity could increase advertisements’ memorability and moderate competing brands’ advertising inference to give an already familiar brand a marketing advantage over a less mature brand [60]. Past research has also demonstrated that brand familiarity could positively influence consumer confidence, brand attitude, and purchase intention [61], aside from determining the effectiveness of advertising [62]. A meta-analysis study also confirmed that brand familiarity could have a positive impact on brand attitude, product types (hedonic vs. mature products), and brand recall [63].

Additional literature further augments the theoretical significance of brand familiarity in relation to prior brand attitude, as it indicates that a consumer’s prior brand attitude can predict both the effectiveness of marketing messages (e.g., [64]) and the consumer’s subsequent attitude and behaviors toward the brand [65,66]. For example, Chattopadhyay and Basu [64] found that when controlling for brand familiarity, a consumer’s prior attitude toward the brand moderated the relationship between message type and message effectiveness to demonstrate the greater effectiveness of a humorous than a non-humorous advertisement. A systematic review of consumer behavior literature also reported that prior attitude was a significant predictor of consumer beliefs and purchase behaviors [67].

Based on the established empirical literature discussed above, a hypothesis is proposed to test the relationship between prior brand familiarity and prior brand attitude.

H5: 

Consumer’s prior brand familiarity is positively associated with their prior brand attitude.

To date, only less than a handful of social media influencer or e-commerce studies have investigated the consumer factors that could help explain the perceived expertise, authenticity, and attractiveness of a spokesperson. For instance, while one study validated that prior brand attitude was a significant antecedent variable for explaining the purchase intention of a consumer reviewer/opinion leader in an e-commerce setting [38], another study confirmed that prior brand attitude was a positive predictor of trustworthiness, expertise, and attractiveness of a spokesperson in an online advertising context [17]. Other than that, only one study had evidenced prior brand attitude as a positive factor for affecting Instagram influencers’ expertise, trustworthiness, and attractiveness [17].

Based on the discussion above and in Section 2.1, it is clear that empirical research remains lacking in terms of exploring the relationship between prior brand attitude and the combination of perceived influencer attractiveness and authenticity factors. Missing the necessary empirical evidence to support a hypothesis to assert the relationship between the antecedent and outcome variables of interest here, the following research questions will be tested to explore these speculated relationships.

RQ1a-b: 

Will prior brand attitude be positively associated with perceived influencer (a) attractiveness, and (b) authenticity?

2.5. Influencer Type and Gender–Product Matchup

As aforementioned, even though virtual influencers can mimic the attractive appearance of and provide content similar to that posted by human influencers, some consumers may be skeptical of virtual influencers’ authenticity because their personas are generated by multiple marketing and engineering teams [68,69]. In particular, social media users may also perceive virtual influencers to be psychologically distant, due to the perceived lack of autonomy associated with these virtual figures, which can lower consumers’ perception of their authenticity [15,69,70]. Even so, as presented in the literature above, while some studies found virtual influencers to be relatively authentic [15,63], other studies presented the opposite result [5,7,27].

In a similar vein, prior studies have also produced mixed findings regarding the perceived attractiveness of a virtual influencer. For instance, study participants reported negative emotions such as uneasiness and the sense of creepiness toward virtual influencers due to their resemblance to an actual human [7,71]. These findings align with the assumption of the uncanny valley hypothesis, which proposes that known non-human entities may elicit a negative reaction from users when they become too human-like [72]. By contrast, another study [16] showed that followers might find virtual influencers attractive because these virtual figures could possess a unique, esthetic, or even idealized visual appearance not bounded by the constraints of the body shapes and physical features of a real human.

Based on the two different bodies of literature explicated here, we will compare perceived authenticity and attractiveness between a human and a virtual influencer via the research questions below.

RQ2a-b: 

Will perceived authenticity and attractiveness differ between a virtual influencer and a human influencer?

Above and beyond filling the empirical gap by comparing the perceived influencer attributes explicated above, whether influencers’ gender could affect their effectiveness in endorsing different types of products is largely unknown. Y.-H. Lee and Yuan [73] demonstrated that the most popular female virtual influencers tended to receive more positive comments, compared to their male virtual counterparts. However, since a product can be regarded as an extension of a consumer’s self-concept, it is possible that consumers who value gender–product congruence may also seek gender cues when reviewing product endorsement messages [74].

To date, only a few studies have investigated the effects of virtual influencers’ endorsement of different product types [7,9]. For instance, Liu and Lee’s [7] work revealed that consumers perceived virtual influencers as less genuine than human influencers across different product types (functional, symbolic, and experimental), making them less effective in product endorsement. In contrast, another study [9] showed that virtual influencers’ recommendations appeared to be more beneficial for utilitarian products, whereas human influencers’ recommendations were more effective for hedonic products.

As limited relevant research has not explored the gender–product matchup in the context of influencer marketing, the following research questions are proposed to test whether different product types endorsed by a male vs. a female human or virtual influencer will have differential effects.

RQ3a-b: 

Will purchase intention differ (a) when a hedonic product is endorsed by a male vs. female human or virtual influencer, and (b) when a utilitarian product is endorsed by a male vs. a female human or virtual influencer?

2.6. Proposed Conceptual Model

To illustrate the interrelations between the theoretical constructs discussed above, a conceptual framework is proposed below (Figure 1).

3. Method

This study adopted a 2 (influencer type: human vs. virtual) × 2 (product type: hedonic vs. utilitarian) × 2 (influencer gender: female vs. male) between-subjects design with random assignments. With advance IRB approval, undergraduate students were recruited from a large university in the Northeastern United States and offered extra credit for their research participation. To estimate the required sample size for the study, we conducted an a priori power analysis using the G*Power formula (Version 3.1.9.7) by setting Cohen’s f-value at 0.25, based on Li et al.’s [6] study on the endorsement effectiveness of virtual vs. human influencers. The power analysis showed that a sample size of 128 is sufficient to achieve the commonly adopted 80% power level [75,76]. We recruited more participants than needed to avoid a potential shortfall caused by participant dropouts and data quality concerns. After removing participants who failed the manipulation check, the experiment yielded 468 valid responses. The final sample size provided sufficient statistical power for our data analysis.

3.1. Stimulus

A mock Instagram account profile was developed for a fictional male and female influencer to represent both a human and a virtual influencer. Each influencer endorsed both a hedonic and a utilitarian product, accompanied by the same comment section. This design approach, similar to that of Li et al. [6], enabled us to eliminate potential confounds such as differences in facial expression, nonverbal gestures, lighting and/or overall aesthetic quality. To indicate whether the influencer is designated as a human or virtual, the mock Instagram post featuring a virtual influencer was labeled explicitly as “virtual influencer” on the “profile page”; the same mock Instagram post showing a human influencer was presented without such a label designation.

The stimulus for each condition promoted either a hedonic or utilitarian product. In particular, the hedonic product condition featured the Airbnb travel rental brand with a photo of the influencer taking a selfie in front of a nature landscape. The utilitarian product condition promoted the NJM insurance brand with a photo of the influencer driving an automobile. For both brand conditions, the brand logo appeared on the top-left corner of the image. The comment in the post highlighted the benefits of the brand and encouraged followers to choose the brand. In total, eight different mock Instagram posts for eight study conditions were shown to the participants; the images of these posts are displayed in Appendix A below.

3.2. Manipulation Checks

To validate the manipulation for influencer type, a t-test was conducted. Results showed that participants in the human influencer condition evaluated the influencer to resemble an actual human (M = 3.35, SD = 2.19) more than a virtual agent (M = 2.40, SD = 1.94; t(463.64) = 4.99, p < 0.001). Participants in the virtual influencer condition rated the influencer to resemble a virtual agent (M = 6.13, SD = 1.45) more than an actual human (M = 5.19, SD = 1.88; t(447.42) = −6.01, p < 0.001).

In terms of product type, participants in the hedonic product condition rated the product as reflecting a greater hedonic (M = 4.24, SD = 1.53) than utilitarian (M = 3.61, SD = 1.71; t(460.32) = 4.17, p < 0.001) value. By comparison, participants in the utilitarian product condition rated the product as embodying a more utilitarian (M = 4.74, SD = 1.49) than hedonic (M = 4.35, SD = 1.42; t(466) = −2.82, p = 0.005) value.

As for influencer gender, participants in the male influencer condition identified the influencer to resemble a male (M = 6.17, SD = 1.05) more than a female (M = 1.32, SD = 0.64; t(387.81) = 60.46, p < 0.001). Contrastingly, participants in the female influencer condition identified the influencer to represent a female (M = 6.59, SD = 0.85) more than a male (M = 1.54, SD = 0.91; t(465) = −62.13, p < 0.001). Based on these results, the stimulus manipulation was deemed successful.

3.3. Procedure

After logging onto the study’s webpage, participants first replied to a set of demographic questions before they were randomly assigned to one of the eight experimental conditions. Following that, they were instructed to view one of the eight mock Instagram posts that featured either a male or female virtual or human influencer before responding to a set of manipulation check questions. Afterwards, they proceeded to complete the post-exposure evaluation measures.

3.4. Measures

All measurement items were evaluated on a seven-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree), unless otherwise indicated. A confirmatory factor analysis (CFA) tested the measurement validity of latent variables. Following that, an exploratory factor analysis analyzed the factor structure of the measurement items for each conceptual dimension, before a Cronbach’s α test tested and confirmed the inter-item reliability for each conceptual dimension. No collinearity concern was present, with the values of the variance inflation factor (VIF) ranging from 1.01 to 1.67 [77].

Perceived influencer similarity measured participants’ perception of social similarity shared between themselves and the influencer in the Instagram post (M = 3.60, SD = 1.49, α = 0.93), using 6 items adopted from Lin and Pierre [40]. Sample items include, “The influencer that I saw (1) is similar to me, (2) is similar to someone from my close circle of friends, and (3) is similar to someone from my larger social network.”

Perceived influencer authenticity described participants’ perception of influencers’ authenticity (M = 4.97, SD = 1.50, α = 0.94), applying 6 items adapted from J. A. Lee and Eastin’s work [29]. Sample items include, “This influencer I reviewed (1) seems kind (2) is good natured (3) is sincere.”

Perceived influencer attractiveness assessed participants’ perception of the influencer’s physical attractiveness (M = 4.12, SD = 1.19, α = 0.88), utilizing a semantic differential scale and 5 items adapted from previous studies [22,78]. Examples of the polarized adjective pairs include “Unattractive/Attractive”, “Not Classy/Classy”, and “Ugly/Beautiful”.

Hedonic consumption motivation was gauged with 6 items, adapted from Voss et al.’s study [45], to indicate experiential motivation (M = 4.94, SD = 1.32, α = 0.93). Example items include “Not fun/Fun”, “Dull/Exciting”, and “Unpleasant/Pleasant”.

Utilitarian consumption motivation was indicated by five items, also adapted from Voss et al. [45], to reflect functional motivation (M = 5.50, SD = 1.24, α = 0.93). Sample items include “Irrational/Rational”, “Senseless/Sensible”, and “Not useful/Useful”.

Purchase intention gauged participant likelihood to purchase the product endorsed by the influencer, utilizing three items (M = 3.01, SD = 1.50, α = 0.90) adopted from Lin et al.’s study [17].

Prior brand attitude evaluated prior participant attitudes toward the two brands under study—Airbnb (M = 5.73, SD = 1.12, α = 0.94) and NJM (M = 3.90, SD = 0.96, α = 0.97)—with 5 items adapted from S.Y. Lee’s study [79].

Prior brand familiarity demonstrated the prior familiarity of the participants with the two brands examined here—Airbnb (M = 5.99, SD = 1.38) and NJM (M = 2.07, SD = 1.79)—with the use of a single measurement item.

The control variables included participants’ gender, age, and annual household income, in addition to the number of human and virtual influencers that the participant followed on Instagram.

4. Result

A confirmatory factor analysis (CFA) was conducted for the data associated with each brand. The CFA results showed a relatively good model fit for the Airbnb model: χ² = 779.85, df = 328, χ²/df = 2.38, p < 0.001, CFI = 0.96, RMSEA = 0.05, TLI = 0.96, NFI = 0.94, IFI = 0.96. The NJM model also had a similarly good model fit: χ² = 942.56, df = 383, χ²/df = 2.46, p < 0.001, CFI = 0.96, RMSEA = 0.06, TLI = 0.96, NFI = 0.94, IFI = 0.96.

Table 1. Results of measurement validity test.

Airbnb	CR	AVE	1	2	3	4	5	6	7
1. Airbnb Attitude	0.94	0.80	0.90
2. Hedonic	0.96	0.83	0.09	0.91
3. Utilitarian	0.93	0.78	0.17 *	0.26 *	0.88
4. Similarity	0.95	0.81	0.09	0.20 *	0.08	0.90
5. Authenticity	0.94	0.75	0.09	0.22 *	0.07	0.44 *	0.86
6. Attractiveness	0.87	0.69	0.08	0.24 *	0.10 *	0.38 *	0.49 *	0.83
7. Purchase	0.90	0.75	0.10 *	0.30 *	0.02	0.38 *	0.49 *	0.37 *	0.87
NJM	CR	AVE	1	2	3	4	5	6	7
1. NJM Attitude	0.97	0.86	0.93
2. Hedonic	0.96	0.83	−0.01	0.91
3. Utilitarian	0.93	0.78	−0.01	0.26 *	0.88
4. Similarity	0.95	0.81	−0.09	0.20 *	0.08	0.90
5. Authenticity	0.94	0.75	−0.16 *	0.22 *	0.07	0.44 *	0.86
6. Attractiveness	0.88	0.64	0.01	0.23 *	0.09	0.36 *	0.48 *	0.80
7. Purchase	0.90	0.75	−0.05	0.30 *	0.02	0.38 *	0.49 *	0.36 *	0.87

Note: * p < 0.05.

shows the results of a measurement validity test (where the diagonal values in bold numbers indicate the square root of average variance extracted for each variable). Specifically, the convergent and discriminant validity for both models (see Table 1) showed that the composite reliability (CR) of all measures exceeded 0.70, the average variance extracted (AVE) exceeded 0.50, and the CR of each measure was greater than its AVE [80].

Testing Results for Hypotheses and Research Questions

The structural equation model (SEM) testing the Airbnb brand, with the inclusion of control variables, showed a relatively good model fit [81]: χ² = 1257.34, df = 614, χ²/df = 2.05, p < 0.001, RMSEA = 0.05, CFI = 0.95, TLI = 0.95, NFI = 0.91, IFI = 0.95 (see Figure 2). Control variables, including gender, age, and household income, as well as the number of human and virtual influencers followed by the participant on Instagram, are not shown in the figures that illustrate the structural equation modeling results.

For the NJM model, with the inclusion of the control variables, the SEM also generated an acceptable model file: χ² = 1341.30, df = 651, χ²/df = 2.06, p < 0.001; RMSEA = 0.05, CFI = 0.95, TLI = 0.95, NFI = 0.91, IFI = 0.95 (see Figure 3).

The SEM results for the Airbnb model supported H1a-b, which found that perceived similarity had a positive effect on (a) perceived influencer attractiveness (β = 0.37, p < 0.001), and (b) perceived influencer authenticity (β = 0.42, p < 0.001). The same is true for the NJM model, where perceived follower–influencer similarity also had a positive effect on (a) perceived influencer attractiveness (β = 0.38, p < 0.001), and (b) perceived influencer authenticity (β = 0.44, p = 0.001).

For the Airbnb model, the SEM results also supported H2a and H2b, which indicated that (a) perceived influencer attractiveness (H2a, β = 0.18, p = 0.001) and (b) perceived influencer authenticity (H2b, β = 0.33, p < 0.001) were each positively related to Airbnb purchase intentions. The same results also held true for the NJM model, as perceived influencer attractiveness (H2a, β = 0.17, p = 0.001) and authenticity (H2b, β = 0.36, p < 0.001) emerged as significant predictors of NJM purchase intention.

H3 proposes that a hedonic consumption motivation will have a positive effect on purchase intention for a hedonic product instead of a utilitarian consumption motivation. SEM results confirmed H3, as hedonic motivation was a positive predictor of purchase intention toward Airbnb (β = 0.16, p = 0.001), instead of utilitarian motivation (β = −0.03, p = 0.594). H4 asserts that a utilitarian consumption motivation will have a positive effect on purchase intentions for a utilitarian product instead of a hedonic consumption motivation. SEM results failed to support H4, as utilitarian motivation (β = −0.03, p = 0.55) did not have any effect on purchase intention toward NJM. Instead, hedonic motivation significantly predicted purchase intentions for NJM (β = 0.16, p = 0.002).

H5 assumes that prior brand familiarity with Airbnb and NJM will be positively and respectively related to prior attitudes toward Airbnb and NJM. SEM results confirmed H5, as prior brand familiarity was a significant predictor of prior attitudes toward Airbnb (β = 0.46, p < 0.001) and NJM (β = 0.48, p < 0.001), in that order.

RQ1a-b explored the relationship between prior brand attitude and perceived influencer (a) attractiveness and (b) authenticity. The results showed that prior attitude toward Airbnb had no effect on perceived influencer (a) attractiveness (β = 0.04, p = 0.389), or (b) authenticity (β = 0.05, p = 0.224). By contrast, prior attitude toward NJM did have a negative effect on (b) influencer authenticity (β = −0.12, p = 0.006) but not (a) influencer attractiveness (β = 0.04, p = 0.379). RQ2a-b examined whether perceived authenticity and attractiveness would differ between a virtual influencer and a human influencer. ANCOVA results showed that there was no significant difference in perceived authenticity (F(1, 453) = 0.29, p = 0.585, η² partial = 0.001) or attractiveness (F(1, 453) = 1.14, p = 0.286, η² partial = 0.003) between a virtual and a human influencer.

RQ3a-b investigated whether consumer purchase intentions would differ due to the matchup between influencer type (virtual vs. human), influencer gender (male vs. female), and product type (hedonic vs. utilitarian). For the hedonic product condition, ANCOVA results revealed that influencer type and influencer gender did not have a main effect on purchase intentions toward Airbnb (influencer type: F(1, 224) = 0.06, p = 0.812, η² partial < 0.001; influencer gender: F(1, 224) = 0.03, p = 0.876, η² partial < 0.001). No significant interaction effect between influencer type and gender was detected in the model [F(1, 224) = 0.004, p = 0.949, η² partial < 0.001].

For the utilitarian product condition, influencer type and gender also did not have a main effect on purchase intentions toward NJM (influencer type: F(1, 224) = 0.86, p = 0.355, η² partial = 0.004; influencer gender: F(1, 224) = 0.143, p = 0.233, η² partial = 0.006) either. No significant interaction between influencer type and gender was discovered in the model [F(1, 224) = 0.13, p = 0.260, η² partial = 0.006].

5. Discussion

The current study aimed to advance our theoretical and empirical understanding of virtual influencers and their impact on consumer behavior, relative to that of human influencers. In particular, we explored the effect of brand familiarity and prior brand attitude on perceived follower–influencer social similarity, as well as the perceived attractiveness and authenticity of influencers. We also tested how influencers’ attractiveness and authenticity impacted hedonic vs. utilitarian brand purchase intentions, alongside a hedonic vs. utilitarian consumption motivation. Structural equation models showed that all the hypothesized paths were validated, excepted for the paths between prior product attitude and perceived influencer attractiveness and/or authenticity measures—depending on the brand—as well as between utilitarian consumption motivation and purchase intention.

Specifically, brand familiarity was a significant predictor of prior brand attitude for both Airbnb and NJM, validating the limited empirical literature addressing the social media influencer context (e.g., [17]). Interestingly, prior attitude toward the NJM brand negatively predicted perceived influencer authenticity and had no relevance to perceived influencer attractiveness. These results are incongruent with that of past work [17]. A plausible explanation for this discrepant evidence may have to do with the fact that NJM does not employ an influencer marketing strategy. The presence of the influencer might have led participants with a more negative attitude toward the brand to appreciate the influencer’s authentic experience expressed in an unpretentious tone, even as participants did not exert a favorable or unfavorable assessment of the influencer’s attractiveness.

Prior attitude toward the Airbnb brand was found to have no effect on either perceived influencer authenticity or attractiveness, contradicting prior research findings [17]. Unlike NJM, Airbnb’s marketing strategy frequently employs both celebrity and non-celebrity influencers. It is possible that participants’ attitudes toward Airbnb did not trigger an evaluation of the familiar authenticity and physical attractiveness typified by an Airbnb influencer seen in social media. By implication, the combined meaning of these findings suggests that depending on the prior consumer attitude toward NJM, an evaluation of the influencer’s authenticity but not their attractiveness may ensue because of the novelty effect. By comparison, regardless of consumers’ prior attitude toward Airbnb, such an attitude may not impact the novel influencer’s authenticity or attractiveness, due to consumer experience with encountering new Airbnb influencers over time.

Contrastingly, perceived follower–influencer social similarity had a positive effect on the perceived attractiveness and authenticity of influencers, consistent with relevant past findings [82]. Unlike the conventional conceptualization of perceived similarity, the current study operationalized and validated perceived social similarity—between followers and influencers in the theoretical context of social identity [40,83]—instead of perceived similarity in product, hobby, and style preferences [33,84]. This finding hence confirms that Gen Z will follow influencers who share a similar social network-oriented social identity, above and beyond shared lifestyle factors [31,85].

As anticipated, the perceived attractiveness and authenticity of influencers were both significant predictors of purchase intention, suggesting that more relatable and likable influencers could potentially help boost favorable consumer intentions to purchase the brand they endorse. These results thus align with past research which indicated that influencer authenticity led to more favorable brand attitudes and purchase intentions [7], in addition to fostering more positive affective reactions [29,86]. By implication, influencers that appear more genuine and more attractive will continue to more effectively benefit the brands that they represent.

Turing to the contribution of consumer motivation to purchase intention, it was hedonic instead of utilitarian consumption motivation that drove purchase intention toward both the hedonic (Airbnb) and utilitarian (NJM) products. This result is similar to past research on travel blog engagement, which reported that although the blog content provides practical information accompanied by enjoyable elements, hedonic motivation was more salient than utilitarian motivation in driving engagement [87]. One possible reason for this outcome may have to do with the fact that NJM’s national advertising campaigns have always adopted a hedonic approach, not unlike other national insurance brands (e.g., Geico and Progressive).

In particular, although NJM promotes itself with the slogan “no jingles or mascots, just great insurance,” its advertising strategy features an array of humorous mascots to enhance its entertainment value and intensify a playful brand image [88]. As social media influencers are typically seen as “cheerleaders” for the brand, their presence is likely to activate a hedonic consumption motivation more often than a utilitarian consumption motivation. Thus, marketers who utilize influencers to endorse a utilitarian product should consider adopting a rational and practical appeal in a light-hearted and fun manner to draw consumer attention, increase brand liking, and elicit a purchase intention.

With regard to perceived authenticity between a virtual and a human influencer, no significance difference was detected. This finding thus both confirms the evidence reported by some prior studies [10,15,68] and disconfirms that of others [5,6,7,27]. The same is true with the results showing a lack of difference in perceived attractiveness between a virtual and human influencer, which again both validates one set of empirical outcomes [16] and invalidates another set of empirical results [7,70]. As Gen Z is the first generation to be using AI tools for school work, recreation, and job functions [89], the social media platforms that they routinely engage are widely populated with AI-generated visual and audio-visual content. For this reason, their acceptance of virtual influencers is not completely unexpected. It is foreseeable that virtual influencer marketing will continue to increase in the U.S., while a steady growth in the virtual marketing trend in Western Europe and East Asia is also anticipated [90,91].

Lastly, the study findings also demonstrated no difference in purchase intention, whether the product was endorsed by a virtual or human influencer or whether the influencer was a male or female. Nonetheless, additional results generated by the ANCOVA test revealed that the female influencer was seen as more attractive than the male influencer. This finding is similar to prior research that showed how female influencers (virtual or human) received more favorable feedback than their male counterparts [72], and they also dominated the influencer market [82]. As Gen Z tends to embrace gender diversity [92], future work could further explore the effectiveness of gender-fluid influencers if such an influencer strategy matches the brand’s identity and image.

In sum, this study revealed that perceived social similarity, a measure reflecting the multicultural social identity among the Gen Z population, had a direct effect on the perceived attractiveness and authenticity of influencers. Perceived influencer attractiveness and authenticity in turn led to a positive effect on purchase intention. As for consumption motivation, a hedonic instead of utilitarian motivation elicited purchase intentions for both hedonic and utilitarian products. These findings thus suggest that when influencer marketing evokes more pleasure and experiential value, it may also be more successful in motivating purchase decision-making regardless of the product type. Since perceived novelty is one of the key drivers for engagement with virtual influencers [5], marketers should consider influencer messaging strategies that embed functional benefits within affectively engaging content when targeting the Gen Z audience for utilitarian products.

5.1. Limitations

The present study has several limitations. First, the Gen Z college student sample was not fully representative of the general Gen Z population. Future work should expand to the broader general Gen Z population to enhance the external validity of our findings, as Gen Z will become the biggest spenders in the market in the next decade. Second, to avoid potential participant bias stemming from awareness or knowledge of real-life influencers, “fake” influencers were used in the study. Utilizing already-known influencers by controlling for preexisting participant bias toward these influencers may increase the ecological validity of the experiment. Third, the current study matched a “fake” influencer with either a real hedonic or utilitarian product. Future research could consider delineating differential effects on marketing outcomes, when comparing a “fake” virtual and human influencer to a real virtual and human influencer endorsing the same or different hedonic or utilitarian products.

Fourth, our study examined how individuals react to the explicit labeling of “virtual” influencers. Additional work could further manipulate an influencer’s visual appearance (e.g., age, facial expression, nonverbal gestures, and/or racial and ethnic characteristics) to provide an expanded understanding of virtual influencers’ effects. Fifth, the current research was a one-shot study conducted with a USA Sample. A preferred methodology could involve a longitudinal study to allow for observing how the use and acceptance of virtual influencers my evolve over time. Lastly, as Gen Alpha is being trained to utilize AI for school work and job functions, it is also essential for researchers to begin learning about how this generation will respond to virtual influencers, as they live, work, and play with AI tools, including collaborative and humanoid robots in the AI age.

5.2. Implications

Historically, the advertising industry has been highly successful in marketing products using non-human spokespersons such as cartoon characters (e.g., the Jolly Green Giant, Keebler’s Elves, Toucan Sam, and Tony the Tiger), animated stuff toys (e.g., Snuggle, Energizer Bunny, and Serta Counting Sheep), animated mascots (e.g., Aflac, Geico, the Pillsbury Doughboy, and the Michelin Man). Hence, if marketers apply their creativity and ingenuity to develop likeable personality traits and welcoming authentic personas for virtual influencers, these human-like replicas could potentially turn into bona fide AI-based advertising spokespersons [93].

It is worth noting that the cost of employing virtual influencers—accompanied by a personality profile and personal history/background—is a fraction of managing celebrity or self-made influencers who could be attached to unanticipated and unpredictable personal conduct problems and reputation concerns. Hence, utilizing virtual influencers is a cost-effective marketing strategy to reach consumers, as long as such a use is congruent with and can enhance the brand’s branding and marketing objectives. Even so, researchers and practitioners alike should carefully identify which product type—hedonic and/or utilitarian, under which product categories and/or sub-categories—may benefit most from employing virtual influencers. For instance, fashion brands, having pioneered the use of virtual influencers, are able to continuously maintain marketing success with their virtual influencer strategy [94].

Furthermore, Gen Z’s preference for short digital video content [95], as evidenced by TikTok’s dominance in attracting 60% of Gen Z users [96], reflects the needs to explore influencer marketing using new video formats. In particular, a prior study showed that the TikTok platform could fulfill users’ self-expression and social interaction needs, which enhances their emotional bonding with influencers and purchase intentions [54]. Live streaming is another video-based format popular in overseas markets such as China, the second largest consumer market in the world [97]. Future work could further compare human vs. virtual influencers’ effectiveness on TikTok and additional social media platforms and in different cultural settings.

In terms of the marketing effectiveness of virtual influencers between cultures, the available literature is very limited. For instance, one study investigated the perceived cultural intelligence, instead of virtual influencer acceptance, between cultures [98]. Another study, which used Hofstede’s framework to classify the cultures of four nation states [99], found that collectivist culture with a low uncertainty avoidance tendency was more accepting of virtual influencers than an individualistic culture with a high uncertainty avoidance tendency [100]. Theoretically speaking, Hofstede’s macro-level monoculturalism model does not address the contemporary multicultural society and transcultural communication environment, where BTS (a K-pop group) can transcend both language barriers and cultural boundaries to become a global brand [39]. Hence, the construct of perceived social similarity, which considers how consumers consider the social similarity between themselves and others in a cross-cultural context [39,40], can provided a theoretically sound basis for exploring the influencer–follower relationship to assess influencer marketing’s effectiveness across cultures.

Lastly, as marketers strategize utilizing virtual influencers’ authenticity and attractiveness to persuade consumers to make product purchase decisions, it is important to consider whether this practice borders on manipulating the consumers with a false reality. This false reality, as presented by virtual influencers that lack autonomy and transparency, is engineered by the social media marketing team. Even though Instagram has a voluntary labeling policy for identifying a virtual influencer, such a policy does not follow the recommended practice by the USA Federal Trade Commission, which states, “If your endorsement is in Instagram Stories, superimpose the disclosure over the picture and make sure viewers have enough time to notice and read it” [101]. It is clear that marketers have a moral responsibility to avoid potentially deceiving consumers through their virtual influencer strategy [69]. This ethical question is a topic that requires continuing research attention at the dawning of the AI age.