Online stars, also called cyberspace stars, web celebrities, online celebrities, or web influencers, are usually focal figures who attract numerous fans by creating and enthusiastically sharing original texts, pictures, or videos on social media (Stever and Lawson, 2013). They gain personal influence and public recognition by frequent interactions with fans, or present desires and wishes that are not fulfilled in practical life to draw massive attention (Zhou et al., 2018). Despite their increasing influence in endorsements, online stars have not been consistently defined across scientific researches. The current study defines “online stars” as self-empowered individuals (Tiggemann and Slater, 2013) who gain fame and recognition by featuring self-generated content and frequently interacting with a large follower base on social media (Wood and Burkhalter, 2014; Schouten et al., 2020), becoming key opinion formers and online broadcasters and hence transforming social capital into cash.

Unlike online stars, social celebrities are often famous actors, singers, artists, supermodels, or athletes, etc., who gain social influence with their professional talent in specific fields and are frequently shown to general public via various media forms (Driessens, 2013). With their positive images in traditional media channels and social interactions, they gain fame and reputation and are greatly favored by certain social groups (Turner, 2004). Although online stars are sometimes also called “online/web celebrities,” the concept of “celebrities” is adopted in a more traditional sense in the current study, referring to individuals who become famous and popular by showing their personal talent and attractiveness, in conventional channels like movies, pictures, commercials and videos. As representatives of certain individual, social and commercial images, they are usually viewed by the public as cultural symbols.

Although celebrities and online stars share similarities, such as being considered to be popular, professional, and trustworthy (Choi and Rifon, 2007; Khamis et al., 2017), the two groups actually differ greatly. Social celebrities normally gain public recognition via traditional and social media, while online stars build their reputation almost exclusively by social media platforms. Further, unlike celebrities who gain fame through contents in various media forms, online stars become self-empowered with their self-generated contents. What’s more, online stars are often “grassroots” people from all walks of life (Russell and Rasolofoarison, 2017). Whereas traditional celebrities seem to be distant, online stars seem more like ordinary people and are perceived as closer to consumers. Hence, social distance, or the perceived psychological distance between the two types of endorsers and their fans, is the key factor differentiating celebrity endorsements and online star endorsements (Stever and Lawson, 2013; Liu and Qin, 2014).

As the predictor of consumers’ buying behavior, purchase intention refers to the likelihood or desire of a customer to buy a certain product (Burke et al., 1985). Numerous factors affect purchase intentions, such as consumer personality characteristics, product internal cues, product external cues, and consumption scenarios, etc. (Dodds et al., 1991). Advertising, a combination of the above factors, has become an effective marketing tool. For both online and offline consumers, once they are exposed to different types of endorsement advertisements, they associate the content with the endorsed products, thereby increasing purchases intention.

Celebrity endorsement is a long-standing and effective driver of consumers’ purchase intention. Celebrities’ public image and fame engender trust in consumers and increase their willingness to purchase the endorsed products (Ayeh, 2015). Moreover, celebrities embody special and symbolic meanings in certain sociocultural environments. Thus, symbolic meaning is injected into the products or brands that they endorse via the process of meaning transfer in consumer purchasing (Knoll and Matthes, 2016). Consumers are also likely to buy products endorsed by celebrities due to the bandwagon and halo effects. In addition, research has demonstrated that advertisements with celebrity endorsements are associated with higher information processing speeds, better product evaluations and increased purchase intention, compared with advertisements lacking such endorsements (Rs and Alawadhi, 2020). Therefore, we propose the following hypothesis:

Online star endorsements represent a new type of endorsement. Similar to conventional celebrities in the sense that both are considered credible, popular and professional, online stars are recruited by more and more companies to present information and recommend purchases to their huge fan bases through impression management, expressiveness and persuasiveness. Online star endorsements are statements made by online celebrities in support of products or brands via advertisements on both traditional and social media. Based on the extant literature on famous persons, we posit that online star endorsements promote consumers’ purchase intention and that the attractiveness or personal charisma of these stars directly influences purchase intention (Park and Lin, 2020). More importantly, the unique qualities of online stars, such as their amiable, topical, and interactive nature, all contribute to consumers’ willingness to buy. Most online stars are considered “common” people and are naturally more amiable and appealing to consumers (Levordashka and Utz, 2016; Jin and Ryu, 2020). Therefore, consumers are more likely to accept online stars’ product recommendations. In public discussions, online stars are topical, and their comments boost interactions, thereby promoting the dissemination of their recommendations among the fans (Elmira and Oxana, 2018). Online stars also use social media frequently to communicate with fans, which further engenders trust from these fans, who unconsciously transfer this trust to the recommended product information (Gong and Li, 2017). Therefore, we propose the following hypothesis:

Self-concept is defined as a person’s cognitive and affective understanding of who and what they are (Joseph, 1982). Individuals have multiple selves but see themselves as a holistic self that is activated and stabilized (Wang et al., 2000). Following most literature on consumer research, we assume that consumers’ selves take the forms of an ideal self and an actual self (Sirgy, 1982). The ideal self is shaped by imagined ideals and goals related to what the person believes that he would like to be or aspire to become, whereas the actual self is based on the person’s perceived self-reality (Sirgy, 1982). When activated, self-concept determines one’s perception of and response to the environment. An individual is more willing to strengthen this self-concept and strive to seek self-enhancement or self-congruence through consuming behaviors, such as buying, presenting, or using certain products that are considered to represent an extension of the self. In this context, self-congruence theory postulates that consumers judge their self-image based on their perception of how closely their self-concept fits with the product’s image. Self-enhancement and self-congruence thus have been considered significant theories in analyses of the connection between consumer self-concept and purchase intention (Ahn et al., 2013).

According to construal level theory, people make abstract mental construals of distal events that differ by the dimensions of psychological distance. Specifically, as psychological distance increases, people construe objects at a higher (more abstract) level. High-level construals, which include the general, superordinate and core features of events, are simpler, more abstract, schematic, and decontextualized representations. In contrast, low-level construals, which include the incidental and subordinate features of events, are relatively more complex, concrete, contextualized, and unstructured representations. Four dimensions of psychological distance are described in construal level theory: temporal distance, social distance, spatial distance, and hypotheticality (Stephan et al., 2011). Despite important differences between these four dimensions, all of them affect the construal level similarly with respect to predicting different psychological states, as demonstrated in the literature (Milkman et al., 2010). Studies have found that, temporal distance triggers different construal levels for people, leading to the production of different selves (Kivetz and Tyler, 2007; Freitas et al., 2008; Rogers and Bazerman, 2008). The social distance generated by an individual’s perception of social objects should have a similar effect. Therefore, we introduce construal level theory to explain why celebrity endorsement and online star endorsement arouse different self-concepts.

When consumers view celebrity endorsements in advertisements, they focus on the celebrities, who are idols in their mind and are perceived to be socially distant. This triggers a high-level construal, motivating the consumer to represent themselves using abstract, core, essential and general features and generating a self-concept that congruent with their own or others’ expectations (Stephan et al., 2011). In this way, celebrity endorsements inspire the ideal self, and consumers strive to maintain a stable and coherent ideal self as a means of self-reflection. As the buying process is an interaction between advertised product information and consumer self-image (Knoll and Matthes, 2016), consumers are likely to change their attitude in response to the influence of celebrity endorsers. That is, activation of the ideal self leads consumers to pay attention to product self-presentation in advertisements. Congruence between product self-presentation and self-image drives consumers to associate the product with their personal image and social status, fostering a closer product–consumer connection (Su and Reynolds, 2017). Consumers who perceive the product as a projection of their self-image and a symbol of the endorsing celebrity buy it to maintain, extend and enhance their ideal selves (Hong and Zinkhan, 2010). Hence, the product is a better match for the ideal self and thus promotes purchase intention. Therefore, we propose the following hypothesis:

Similarly, when consumers view online star endorsements in advertisements, they focus on the online stars, who are perceived to be common persons or even friends. This perception of natural social closeness results in a low-level construal, diverting consumer s’ attention from inner features. Thus, consumers represent themselves using concrete, superficial, subordinate and contextualized psychological features and thus generating a self-concept at close range (Qian and Park, 2021). In this way, online star endorsements activate the actual self, leading consumers to pay more attention to the actual functional value of the product, instead of its symbolic value. Strengthened as signifiers, online stars are more closely linked with the recommended products than with idols or dreams (Khamis et al., 2017). To speed up the spread of information, these stars usually aim to facilitate the understanding of product information and thus highlight a product’s functional value. In this case, consumers are likely to buy products endorsed by online stars to achieve a congruent self-image and extend the actual self (Sirgy, 1982). Therefore, we propose the following hypothesis:

Although social media platforms, such as Sino Weibo, WeChat, and Kuaishou, provide new opportunities for celebrities to promote themselves and interact with fans, the mysteriousness, distance, and idolatry associated with these celebrities remain unaffected. They remain symbolic as distant idols and dreams that are not readily approachable. The social distance between consumers and celebrities leads to activation of the ideal self, which represents consumers’ expectations and ambition. As a result, consumers who purchase the product are willing to trade some functional value, or even precious resources like time and money, in exchange for rewards, such as increased expression of their personality, identity and image. In advertisements, symbolic appeal means more emphasis on representing symbolic meaning for consumers, on the basis of creating an emotional state of mind for consumers (Johar and Sirgy, 1991; Choi et al., 2012). As symbolic appeal in advertising is anchored on creating ideal goals and images and delivering a certain lifestyle, social status, and signified meaning via products (Wolter et al., 2016), it is a better match for consumers whose ideal self is activated by celebrity endorsements. Such consumers are more likely to purchase products endorsed by celebrities to achieve a more congruent and better extended ideal self (Wang and Scheinbaum, 2018). Therefore, we propose the following hypothesis:

Online stars use social media platforms to release pictures, texts and videos that represent real self-images and to interact frequently with their followers. As noted previously, online stars are perceived to be ordinary people, which reduces their social distance from consumers. These interactions are more likely to activate the actual self. To guide action, such consumers who purchase a product are more likely to sacrifice internal value for increased instrumental and practical value. Functional appeal in an advertisement involves informing consumers of the main functions, outstanding performance and comparative advantages of the product via concrete, peripheral, and functional messages (Johar and Sirgy, 1991; Ashley and Tuten, 2015). In this case, consumers’ actual self is triggered by online star endorsements, resulting in a preference for functional appeal in advertisements. Such consumers are more likely to purchase products endorsed by online stars to achieve a more congruent and better extended actual self (Ahn et al., 2013). Therefore, we propose the following hypothesis:

We conducted four studies to test our hypotheses. In Study 1, we examined the influences of celebrity and online star endorsements on consumers’ purchase intention and determined that both associations were positive. In Studies 2 and 3, we explored these positive associations and determined that they are mediated by consumers’ ideal self and actual self, respectively. Finally, in Study 4, we showed that this mediation effect is conditional upon the type of advertising appeals.

To measure the constructs in our studies, we used or adapted established scales with reference to prior research (see Table 1). A confirmatory factor analysis (CFA) including all study constructs produced a satisfactory overall fit. Factor loadings (more than 0.700), Cronbach’s alphas (over 0.700), CRs (greater than 0.700), and AVEs (ranging from 0. 529 to 0. 783) all point to high levels of reliability and convergent validity of the measurements (Table 1). In addition, the square root of the AVE value is higher than the correlation coefficients for all latent variables, indicating that high discriminant validity.

We conducted a pilot study to determine the appropriate stimuli for formal experiments. To eliminate confounding variables and ensure good validity and reliability, we used real celebrities and online stars as the experimental dataset. We also drew on authoritative selection methods developed by scholars, such as Sengupta et al. (1997) and Biswas et al. (2006) to ensure that suitable stimuli were selected for this study.

First, we selected all celebrities who appeared in all rankings, such as Baidu Billboard, Tencent Billboard, and Weibo Billboard, ranked them in order, and finally identified 10 celebrities of different genders, excluding those who had received negative publicity. Second, using the same selection method, we listed 20 online stars who appeared in all rankings, such as China online stars big data report, listed them in general order and finally selected 10 online stars of different genders. Third, we invited 55 undergraduates (M_age = 21.257 years; 30 female) at a comprehensive university in Wuhan to complete a questionnaire. In this questionnaire, we listed the names of the 10 stars and 10 online stars and attached their photos, which were modified using Photoshop to ensure that they were identical in size, brightness, pixels, and colors. To rule out effects of popularity, expertise, trustworthiness and attractiveness, we introduced the widely used celebrity trait scale developed by Ohanian (1991) and Ding et al. (2005). Specifically, subjects rated on the popularity, expertise, trustworthiness and attractiveness of the celebrities and online stars on 7-point semantic differential scales. For example, expertise was measured by four items with 7-point scales: expert/not an expert, experienced/inexperienced, qualified/unqualified, skilled/unskilled.

Using data from 52 valid questionnaires (those that lacked key information or were answered sloppily were ruled out), we selected Zhao Liying and Papi Jiang, who received the highest scores, to represent female celebrities and online stars, respectively. As shown in Table 2, we found no significant difference between the respondents’ perceptions of these stimuli in terms of popularity, expertise, trustworthiness, and attractiveness. We further selected Yang Yang and Liu Yu, who also received the highest scores, to represent male celebrities and online stars, respectively. Similarly, we found no significant differences between the respondents’ perception of the common features of the celebrity and online star (Table 2). To better allow the participants to distinguish between celebrities and online stars in the following formal experiments, we added a text description of each to the stimulus materials, drawing on the research methods of Atkin and Block (1983).

After setting our research stimuli, we processed photos of the celebrities, online stars and products as described earlier, and used Photoshop to combine these photos into single images. In each image, the celebrity or online star is shown on the left side and the product is shown in a relatively smaller shape on the right side.

In Study 1, we explored the effects of celebrity endorsements and online star endorsements on consumers’ purchase intention. We selected a facial cleanser endorsed both by celebrities and by online stars, using a fictitious brand X to rule out any effect of brand association.

The main purpose of Study 2 was to verify that celebrity and online star endorsements activate consumers’ ideal self or actual self, respectively, thereby enhancing their purchase intention. Specific stimulus information for study 2 is available in Supplementary Appendix.

In this study, we followed the approach developed by Kettle and Häubl (2011) to activate the ideal and actual selves, by using the situation of drinking beer in a public or private location as the stimulus (Graeff, 1997), which is an established way to activate self-concept.

We proved in Study 2 that celebrity endorsements activate consumers’ ideal self, whereas online star endorsements activate the actual self, both of which promote purchase intention. We then assume that, exposed to celebrity endorsements, consumers with activated ideal self are more willing to buy products endorsed by celebrities. Similarly, after exposure to online star endorsements, consumers with activated actual self are more likely to find online star-endorsing products attractive. To further test our assumptions, we conducted Study 3.

In this study, to examine the conditional factors for the effects of celebrity endorsements and online star endorsements on purchase intention, we introduced symbolic appeal and functional appeal in advertisements as moderating variables. We then sought appropriate types of verbal descriptions in advertisements. We conducted this study to test Hypotheses 5 and 6.

The current research is among the first to precisely define the concept of online stars and differentiate it from the concept of celebrities in a traditional sense. As most prior studies focus on the effects of celebrity endorsements (Atkin and Block, 1983; Biswas et al., 2006), the influence of online star endorsements is rather understudied due to limited differentiation between the two groups. Though online stars are distinct from celebrities in the sense that they are socially much closer to consumers, our research suggests that celebrity and online star endorsements are not significantly different in their ability to increase consumers’ purchase intention. Seemingly incongruent with Schouten et al. (2020)’s findings, where significant difference in trustworthiness was found between online stars and celebrities, our study shows that both groups influence purchase intention likewise, as we understand online stars as similarly trustworthy to traditional celebrities since they are both famous to the public.

A second contribution is the importance of exploring the mechanism underlying the effects of online star endorsements and celebrity endorsements. Based on the theories of self and construal level theory, our results show that self-concept mediates the influence of online star endorsements and celebrity endorsements on purchase intention. While extant literature on endorsement effects seem rather confined to investigating the expertise, popularity, trustworthiness, and attractiveness of the endorsers (Park and Lin, 2020; Schouten et al., 2020), our research provides a new perspective by examining consumers’ self-concept in different endorsing situations, supplementing the literature on self-concept and consumer purchase intention (Oumlil and Erdem, 1997).

Third, instead of considering emotional/rational appeal as boundary conditions like previous endorsement studies did (Ruiz and Sicilia, 2004; Andreu et al., 2015), we have found a better moderator using symbolic/functional appeal when investigating celebrity/online star endorsements effects. By examining the moderating effect of advertising appeals, we have gained a better understanding of the situations wherein celebrity endorsements and online star endorsements increase purchase intention.

Our research has important practical implications for advertising and product marketing researches. When considering endorsers, managers may select both celebrities and online stars in accordance with current marketing practice (Milkman et al., 2010), as they influence purchase intention in a similarly significant way. For example, Mercedes-Benz selected traditional celebrities and a popular streamer to endorse its products.

Second, our results also suggest managers consider the significance of appropriately matching the advertising appeal with the endorser type. Specifically, companies may adopt symbolic appeal when a celebrity endorses their product, reinforcing consumers’ ideal self and promoting purchase intention. Similarly, using functional appeal to match online star endorsers, companies activate consumers’ actual self and drive them to buy.

Third, enterprises should consider promoting purchase intention by activating consumer self-concept, which is dynamic and malleable (Freitas et al., 2008; Hong and Zinkhan, 2010). Different activation of consumers’ self-concept increases their willingness to buy products endorsed by either celebrities or online stars to maintain a congruent self. Therefore, celebrity-endorsed products might describe themselves for benefits, such as personal charisma or perfect image in public places, to activate the ideal self. Likewise, companies using online star endorsements to activate the actual self, might encourage consumers to purchase by providing information on product core functions to meet their diverse needs.

This research also has some limitations. First, research on online stars remains at a nascent stage. Although we included celebrities and online stars of both sexes in our studies, we failed to include online stars of an older age because most of them are between 18 and 35. Future research might examine the influence of celebrity endorsers and online star endorsers of different ages on consumers’ purchase intention as the pool of online stars expands. Second, this research only validates the interaction of advertising appeals with endorsement types. Future research could examine other possible moderating variables, such as consumer cultural factors, consumer involvement, communication channels and consumption contexts. Third, this research only measures the independent effects of celebrity or online star endorsers. Future studies could investigate the influence of using a combination of celebrity and online star endorsements on consumers’ purchase intention.