“Airbnb Be Like…”: Examining the Impact of Memetic Doppelgängers on Meme Virality and Brand Image

Abstract

This article examines how internet memes, a distinctive type of user-generated content, have served as a vehicle for the co-destruction of the Airbnb brand. For this purpose, we discuss the relevance of internet memes alongside the doppelgänger brand image concept, which explains the logic behind disparaging brand imagery that circulates within internet culture. A comprehensive research model is developed that links several meme- and user-related characteristics with both meme virality and the brand image of Airbnb. The proposed multilevel model is tested using a split-sample approach on a dataset that includes 121 Airbnb-related memes evaluated by 3,664 participants. The results reveal that the virality of disparaging memes is mainly driven by humor, unambiguity, pictorial format, inoffensive content, and “active” user status. The perceived brand image is, however, mainly affected by the valence (negativity) of the memes, while the meme content and user status are also important.

Introduction

The emergence of Airbnb, a company synonymous with the sharing economy, represented a significant disruption to the tourism sector (Guttentag et al., 2017). Its immense growth and disruptive business model made Airbnb “the most important company of its type,” an indispensable tool in travel planning (Yang et al., 2019), and “the primary frame” for investigating the relationships between the sharing economy and tourism (Hall et al., 2022). Airbnb has developed from the initial romanticized notions of peers “sharing their beds” into a powerful global brand (Menapace, 2019). Airbnb built its brand by emphasizing “its embrace of a community of people, not spaces” (Botsman & Capelin, 2016, p. 3), while also providing an open platform that enables users to engage in brand co-creation.

However, after the initial honeymoon period, cracks began to appear in the relationship between Airbnb and its users, with the platform’s success being tainted by various complaints and controversies (see Huertas et al., 2021). A wave of negative incidents related to Airbnb has led users to share their horror stories on social media which started multiplying to the point of becoming an internet meme (Armitage, 2015). Another controversy erupted after Airbnb’s introduction of the new logo called Bello, which was aimed at becoming a universal symbol of “sharing and belonging” and representing Airbnb as a “warm company that stands for people, places, and love” (Andrivet, 2021). As part of the campaign, Airbnb also developed a tool called Create. It enables users to customize the Airbnb logo, but the tool backfired as users started to submit their own doppelgänger logo creations in the form of internet memes (see Figure 1a) and pointed out a number of alternative (mostly sexual and derogatory) meanings that the new logo seemed to represent. Eventually, Airbnb could no longer hold the narrative, with the mainstream media (e.g., CNN, Guardian, and Forbes) reporting that the new “obscene” logo is a “creative misfire” that sparked a “wave of ridicule” and resulted in “social media embarrassment” (Wainwright, 2014).

Figure 1.

Examples of Airbnb-related memes: (a) Airbnb logo meme, (b) Where’s Johnny meme, (c ) Sophisticated cat meme, and (d) Matrix meme.

These events demonstrate that Airbnb represents a relevant study setting, not only as a peer-to-peer platform, but also as a co-created brand suffering from destructive consumer responses (Gebauer et al., 2013; Kristal et al., 2018; Lund et al., 2019). Over time, such anti-branding efforts can blend into a coherent set of opposing meanings that have serious detrimental effects (Cova & White, 2010; Giesler, 2012; Lund et al., 2019). Disparaging brand messages undermine a brand’s perceived authenticity and can result in brand avoidance, negative WOM, brand hate, consumer boycotts, declining sales, and even declining stock prices (Bucher et al., 2017; Davis et al., 2016; Kucuk, 2019a; Morhart et al., 2015; Pruitt & Friedman, 1986). The recent “meme-stock” phenomenon, for instance, clearly demonstrates the power of collective user efforts to cause billion-dollar damage (Lipschultz, 2024).

The circulation of Airbnb’s doppelgänger images also shows that visually oriented social media is becoming increasingly popular and relevant for the analysis of brand-related user-generated content (UGC) (Nanne et al., 2020). While visual content, such as photographs, has always been an important aspect of tourism, “digital visual turn” corresponds with new media types and increases potentials for instantaneous sharing and manipulation of UGC (filtering, remixing, curating) and requires new managerial and research approaches (Fatanti & Suyadnya, 2015; Gretzel, 2017; Minazzi, 2015).

Since different types of UGC function in different ways (Li et al., 2023), it is not sufficient for tourism research to examine only textual content and tourist reviews, as it should also consider any platform and form of UGC that spreads tourism-related content. In today’s hyperconnected world and era of digital dis/intermediation, where different stakeholders co-create brands in an open, bottom-up manner, traditional concepts and models may be inadequate to understand how UGC influences individuals’ perceptions and actions when interacting with such content (Gielens & Steenkamp, 2019; Lund et al., 2019; Pitt et al., 2006; Swaminathan et al., 2020; Tierney et al., 2016). Hence, considering the controversial developments of the Airbnb brand and the lack of theoretical underpinnings in the UGC literature, an important research question is how memes as a particular UGC form represent tourism brands, such as Airbnb. In particular, there is scant knowledge concerning (1) the impact of memes as a visual and unconventional “review” type of UGC, in which both the memes’ and the users’ characteristics need to be considered; (2) the negative aspects of a tourism brand image (i.e., Airbnb) on social media; and (3) the subsequent actions of users, such as viral sharing of the content.

We address these research gaps by theoretically advancing and empirically investigating the issue of co-created memetic disapproval of Airbnb on social media. More specifically, the aim of this study is to examine the perceptions and related “alternative” (critical, playful, humorous) memetic representations and evaluations of Airbnb, which complement those that are company-projected or captured in conventional (textual) tourist reviews. To accomplish this, we rely on the concept of the doppelgänger brand image to explain when and why critical brand imagery circulates within popular culture, as well as what its strategic implications are (Thompson et al., 2006). Here, the doppelgänger concept is applied to explore brand image as an idiosyncratic, cultural, and communal phenomenon driven by the agency of engaged actors and stakeholders. A doppelgänger brand image is defined as a “family of disparaging images and meanings about a brand that circulate throughout popular culture” (Thompson et al., 2006, p. 50). As such, consideration of doppelgänger imagery can (1) shed light on why disparaging brand imagery circulates within engaged communities and the broader internet, (2) explain how such imagery exposes the “true nature” of certain popular brands and undermines their authenticity, and (3) provide strategically valuable insights and implications (see also, Chhabra, 2020; Giesler, 2012). While highly relevant, doppelgänger imagery has so far been explored only in a handful of studies, of which the most comprehensive were qualitative, exploratory, and based on small samples (Geisler, 2012; Thompson et al., 2006). Given that these studies did not focus on memes and their virality, there is a lack of knowledge that could explain the spread of doppelgänger imagery on social media.

To provide a theoretical underpinning for our study of user responses to memetic doppelgängers, we integrate a classic brand information processing framework (Maclnnis & Jaworski, 1989) with the contemporary meme virality framework (Malodia et al., 2022). Building on the theoretical integration of these two frameworks and the literature on UGC, we develop a conceptual model that links the memetic representations of Airbnb with subsequent individual perceptions and actions while controlling for individual characteristics. In the methodology section, we employ a split-sample approach to test the proposed (multilevel) model on a dataset comprising 121 Airbnb-related memes evaluated by 3,664 participants. We conclude with a discussion of the results and suggestions for future research.

Our contributions to the sharing economy and tourism research literature are threefold. First, we integrate the doppelgänger brand concept and meme virality framework as novel theoretical elements of a more general brand information processing framework (Maclnnis & Jaworski, 1989). This theoretical integration enables us to explain the logic and effects of negative co-branding and consumer engagement in brand devaluation. Second, our discussion and empirical examination of memes reveal how this distinctive type of UGC can serve as a vehicle for the viral co-destruction of established brands, such as Airbnb. Third, our comprehensive research model and multilevel analysis, which illuminate the specific effects of diverse meme- and user-related characteristics, serve as an example of more holistic, rigorous, and valid research into the effects of memes, an area previously dominated by exploratory research designs.

Literature Review

Internet Memes as a Distinctive Type of Tourism-Relevant UGC

Internet memes are defined as digital items with common characteristics that are imitated and reiterated around the web (Nissenbaum & Shifman, 2018). They act as units of popular culture and, during circulation, create a shared cultural experience among users (Shifman, 2013). While memes also address and evaluate tourist brands, tourism literature tends to focus on a different type (genre) of UGC, namely on (textual) tourist reviews (see Munar, 2010, 2011).

Tourist reviews are commonly found on reviewing platforms, such as TripAdvisor, and booking platforms, such as Airbnb. Such platforms have strict rules which ensure that the reviews posted are realistic evaluations, based on “actual, specific, and first-hand (original) experiences,” from known authors. In the case of Airbnb, tourist reviews explicitly evaluate hosts/places, but not the Airbnb brand itself (business model, reputation, etc.). Tourist reviews on sharing economy platforms—Airbnb in particular—are typically heavily positively skewed (Meijerink & Schoenmakers, 2020).

Memes differ from typical tourist reviews in their (visual) form and humorous style, but also in additional characteristics and more critical (negatively biased) stance, all of which make them stand out from other types of “reviews.” Memes are not meant to be realistic as they focus mostly on the message they are trying to convey (Börzsei, 2013). They are not original (first-hand) creations, but imitations (e.g., doppelgänger images). They are also anonymous and flourish on “grassroots” social media (e.g., 9GAG, 4Chan, Tumblr, Reddit), where moderation is less strict and creative freedom is allowed. Since memes are collectively co-created, and understanding pop culture references is essential for understanding them, memes cannot be examined outside of the cultural context (Tanni, 2020).

This also holds true for all meme examples ¹ in Figure 1. The meaning of Airbnb’s “Logo” meme (Figure 1a) is for instance clearer when one understands the history of Airbnb’s “social media embarrassment” (see Wainwright, 2014). The second meme, “Here’s Johnny!” (Figure 1b) is based on a scene from The Shining, which has been deemed “officially the scariest” movie of all time (Child, 2013) and clearly evokes a “scary” image of Airbnb. The third example (in Figure 1c) is known as the “Sophisticated cat” meme and is open to different interpretations. One of them could be that becoming an Airbnb host in New York City is a good business idea due to the high demand for short-term rentals in the area. Alternatively, the cat in a business suit could suggest that Airbnb is a symbol of a new “predatory laissez-faire platform” of capitalism (O’Regan & Choe, 2017). The fourth example (Figure 1d) is, however, much clearer with regard to the intended meaning and message about Airbnb’s negative impact on housing and communities. Each meme can, thus, convey a variety of meanings and serve a variety of purposes, among which (co-)destructive motives, such as provocation, punishment, mocking, protesting, and anti-corporate brand critique, are frequent and prominent (Brubaker et al., 2018; Davis et al., 2016; Shifman, 2014).

Memes are thus useful for understanding users’ negative opinions in the case of Airbnb. Due to the distinctiveness of memes, it is expected that they can illuminate those aspects and mechanisms of brand co-destruction that are not inherent in conventional tourist reviews and would otherwise remain overlooked. In this regard, internet memes are a neglected type of UGC in tourism, which is relevant due to the fact that memes “frame and reflect a deeper social mindset in regard to important issues” (Shifman, 2019).

Studies about memes have recently also emerged in tourism literature. They addressed media-induced tourism (Yhee et al., 2021), destination image representation (Kolar & Walanchalee, 2020), tourism influencers (Zhang & Huang, 2022), and travel-related memes shared during the COVID-19 pandemic (Pabel & Turnšek, 2022). However, there is a notable lack of studies on how memes criticize and co-destruct tourist brands. A possible reason for this could be the prevailing focus of previous studies on textual tourist reviews and the fact that memes are conceptually and semantically a complex type (genre) of UGC (see Cancelas-Ouviña, 2021) that pose significant research challenges, which is why they are also referred to as “conceptual troublemakers” (Shifman, 2014).

Conceptually, memes exhibit several typical characteristics, such as humor, polysemy, intertextuality, content/theme, target, form, and stance (Malodia et al., 2022; Shifman, 2014). The perceived humor of a meme refers to users’ assessments of the extent to which a meme is funny and amusing (c.f. Warren et al., 2019, p. 1026). In contrast, polysemy is conceptualized as the existence of at least two distinct interpretations of the same meme (c.f. Puntoni et al., 2010, p. 52). Intertextuality implies that memes include cross-references to different persons, items, or events from popular culture (Laineste & Voolaid, 2016, p. 32). Content and targets refer to identifiable subject matters or persons within a family of memes on a particular topic (Harvey et al., 2019). Lastly, meme stance is the expression of a positive, negative, or neutral opinion toward the subject matter of a meme (Dynel & Poppi, 2023). These characteristics are of central importance for the creation and diffusion of memes, which is a complex, messy, and dynamic process. Memes are continuously being created, meaning that “only the fittest memes survive” (Bernstein et al., 2021, p. 55) while those that fail to capture sufficient attention or to emotionally resonate within a community quickly vanish (Miltner, 2018).

The literature examining the impact of specific conceptual characteristics of memes on their virality is limited and fragmented. Typically, each study focuses on only one or two meme characteristics. For instance, Malodia et al. (2022) found that consumers are more willing to share humorous memes, while Ling et al. (2021) observed that memes that include images of human faces are more likely to go viral than those featuring other images. Guadagno et al. (2013) reported that consumers are more likely to share memes that evoke positive emotions. However, none of these studies examined the simultaneous effects of the previously defined characteristics of memes on their virality. Moreover, the literature examining the effects of memes on brand image is even more limited.

Doppelgänger Brand Imagery

The sharing economy, hyper-connectivity, and new branded entities, such as customer-to-customer platforms, challenge branding and cause the “broadening” and “blurring” of brand functions and roles. As a result, a broader conceptualization of shared brands needs to acknowledge that they also function as catalysts for social interaction and arbiters of controversy, leading to a higher inherent risk of brand crises (Swaminathan et al., 2020). Here, the concept of doppelgänger brand image becomes particularly relevant since it addresses the tenets and risks of co-created brands and the implications of contested brand meanings caused by co-destructive users. The term “doppelgänger” means “double-walker” and refers to a critical version of the original brand image, thus representing brand twins that are not desired from the company’s perspective. The concept focuses on cultural intermediaries and examines the co-creation (and contestation) of brand meanings within the context of popular culture. One of the key premises of cultural branding is that if brands become powerful icons, they can also experience a process of de-iconization. In this manner, an emotional branding strategy (embraced by Airbnb) is more likely to expose brands to cultural backlash than a benefit-based strategy, and this spurs the creation of co-destructive doppelgänger imagery (Thompson et al., 2006).

In this regard, the concept of doppelgänger branding overlaps with related concepts such as brand assemblage (Rokka & Canniford, 2016), brand remake, brand hijacking, brand attack, and brand co-destruction. These all represent “non-collaborative” forms of brand co-creation (Kristal et al., 2018; Siano et al., 2022). Co-destruction is, however, a broader term, typically used with reference to value (Buhalis et al., 2020; Guan et al., 2020), and textual tourist reviews (Dolan et al., 2019; Sthapit & Björk, 2019). Accordingly, doppelgänger seems to be a more precise and suitable concept for the study of unwanted brand imagery. Some recent doppelgänger studies that explore memes (see Chhabra, 2020; Kotwal, 2019; Popp et al., 2016) demonstrate that, today, memes are a relevant vehicle that effectively convey doppelgänger brand imagery.

Theoretical Background

While various theories and frameworks have been introduced and employed to shed light on how individuals respond to marketing communications (for example, see Choe et al., 2016; Duncan & Moriarty, 1998; Meyers-Levy & Malaviya, 1999), the integrative framework of brand information processing from advertisements (Maclnnis & Jaworski, 1989) provides a valuable starting point for understanding how individuals perceive and respond to memes as an “informative stimulus.” The cornerstones of the original theoretical framework are the antecedents (i.e., communication stimuli), brand information processing, and its consequences (i.e., consumer reactions). More specifically, the theory predicts that exposure to an advertising stimulus interacts with consumer characteristics and leads to the processing of brand information, which ultimately results in cognitive, emotional, and attitudinal responses (Maclnnis & Jaworski, 1989). This framework has been successfully employed to explain brands’ relationships with social media influencers (Aw & Chuah, 2021), responses to augmented reality (Yim et al., 2017), consumer responses to social media communications (Schivinski & Dabrowski, 2016), and their social media involvement in tourism (Leung & Bai, 2013).

Since consumers’ perceptions of brand doppelgänger imagery have not been examined in previous qualitative studies (Freund & Jacobi, 2013; Giesler, 2012; Hietanen et al., 2018; Thompson et al., 2006), we see merit in elaborating the information processing framework by including two universal dimensions of brand image, namely warmth and competence (Kervyn et al., 2012, 2021). A brand’s warmth image encompasses consumers’ perceptions of a brand’s “cooperative and altruistic or competitive and exploitative intentions,” and a brand’s competence image is defined as consumers’ perceptions of the brand’s “ability to act upon these intentions” (Halkias, 2022, p. 2). Warmth image is reflected in traits such as “helpfulness, sincerity, friendliness, and trustworthiness,” while competence image is reflected in “efficiency, intelligence, conscientiousness, and skill” (Kervyn et al., 2012, p. 167).

We propose that these two dimensions represent key cognitive responses to the brand-related information that a meme conveys (Maclnnis & Jaworski, 1989). We base this proposition on several arguments. First, warmth and competence have been seen as fundamental dimensions of social cognition (Fiske et al., 2007), which means that they are both cornerstones of numerous theoretical models of social perception (Abele et al., 2020). Second, scholars from various disciplines identify warmth and competence as the main dimensions along which people form perceptions of various agents, such as individuals, social groups, animals, and brands (Gidaković et al., 2022). Third, the extant literature provides evidence that various aspects of communication, such as advertising humor (Hoang et al., 2023; Peter & Ponzi, 2018), communicator power (Dubois et al., 2016), and writing style (Hwang et al., 2022), may influence consumers’ evaluations of both warmth and competence. Accordingly, we subsequently hypothesize that Airbnb-related memes contribute to the creation of doppelgänger brand imagery that impacts consumers’ perceptions of Airbnb’s warmth and competence.

Since MacInnis and Jaworski’s (1989) integrative framework does not account for behavioral reactions, such as content sharing, and solely considers consumer characteristics as antecedents, we see an opportunity to integrate it with the meme virality framework proposed by Malodia et al. (2022). In the latter framework, the antecedents of meme virality consist of content-, customer-, and media-related factors. Malodia et al. (2022) found that the content-related factors (relevance, iconicity, humor, and spreadability) have a significant effect on the virality of memes. We thus acknowledge that memes need to go viral for doppelgänger brand imagery to be effective and practically relevant, which is why we also investigate the effects of Airbnb-related memes on meme virality.

In the subsequent sections, we integrate predictors related to meme user characteristics and brand-related cognitive responses (i.e., brand image) from Maclnnis and Jaworski (1989) with predictors related to meme stimulus characteristics and behavioral responses (i.e., meme sharing) from Malodia et al. (2022). Based on these outlines, we proceed with the development of our conceptual model and proposed hypotheses.

Conceptual Model and Hypotheses Development

Previous studies on memes were predominantly focused on isolated aspects and content, such as content themes (Kolar & Walanchalee, 2020), specific “families” of memes (Brubaker et al., 2018), particular styles of humor deployed (Taecharungroj & Nueangjamnong, 2015), or various consumer- and media-related factors (Malodia et al., 2022). Therefore, few studies have holistically examined how the conceptual characteristics of disparaging memes impact meme virality while controlling for content- and consumer-related factors (Berger & Milkman, 2012). Thus, we focus our hypotheses on four distinguishing conceptual characteristics of memes: perceived humor, perceived negativity (i.e., stance), polysemy, and intertextuality. Moreover, in line with the brand information processing model, we include brand-related dimensions—warmth and competence—as the outcomes of meme characteristics. As discussed in more detail below, we test our conceptual model on a dataset that combines a broad range of Airbnb-related memes that were evaluated by a large sample of participants. We include several control variables that capture the heterogeneity of the memes in our sample: meme content theme, meme form, and meme target. To control for heterogeneity among the study participants, we control for their past experiences with memes and Airbnb as well as their demographic characteristics (age, gender, income, and education). The conceptual model is shown in Figure 2.

Figure 2.

Conceptual model.

Humor is probably the most frequently explored driver of virality. Given that humor is integral to virality (Jenkins et al., 2013), humorous memes are more likely to become popular (Lankshear & Knobel, 2019; Malodia et al., 2022; Shifman, 2013). Various motivations, such as personal involvement in a situation, creative self-realization, resistance to authority, and self-presentation, have been proposed as drivers of the virality of humorous content (Jensen et al., 2020). Similar findings are reported in tourism literature, as Ge and Gretzel (2018) found that humor is a significant predictor of engagement among firm-initiated social media posts. Therefore, we predict the following:

• H_1a: The perceived humor of a meme has a positive effect on meme virality.

Humor is a frequently used impression management tool, as it can signal intelligence, which is a trait associated with a brand’s competence image (Bitterly & Schweitzer, 2019). Humor also functions as “a social lubricant that facilitates interpersonal connection” (Béal & Grégoire, 2021, p. 244) and can thus improve perceptions of warmth (Bitterly & Schweitzer, 2019). Research from psychology, marketing, and organizational behavior shows that humorous messages positively impact receivers’ perceptions of warmth and competence. For example,DiDonato et al. (2013) found that perceived humor predicts warmth and competence of romantic partners, and Hoang et al. (2023) showed that when advertisements include humorous incongruity, they improve the brand’s competence image. Humor was found to have a positive effect on brand image when studying memes (Teng et al., 2022). Similarly, Brender-Ilan and Reizer (2021) found that when managers use humor in their emails to employees, they benefit from an improved warmth image. In the context of interpersonal communication, Bitterly and Schweitzer (2019, p. 84) reported “a robust and positive relationship between the use of humor and perceptions of both warmth and competence.” Therefore, we predict the following:

• H_1b: The perceived humor of a meme has a positive effect on warmth image.

• H_1c: The perceived humor of a meme has a positive effect on competence image.

Since doppelgänger brand imagery has been conceptualized as a negative (i.e., disparaging) phenomenon (Giesler, 2012; Thompson et al., 2006), we focus only on negative and ambiguous/neutral Airbnb-related memes. Therefore, we operationalize a meme’s stance with the construct of perceived negativity, which we define as the degree to which consumers perceive a meme to “impugn the character, record, or positions” of its target (Pinkleton et al., 2002, p. 14). Advertising scholars distinguish between humor and perceived negativity in marketing communications (Thomas & Fowler, 2021; Warren et al., 2019), and we adopt the same approach since it is considered relevant for brand image examination. Positivity is the key factor to meme virality (Shifman, 2014), and studies from non-memetic contexts (e.g., YouTube ads, online articles, tweets) also demonstrate that positively valanced content is more viral than negative content (Berger & Milkman, 2012; Han et al., 2020; Tellis et al., 2019). For instance, Eckler and Bolls (2011) reported that negatively toned (versus positively toned) online ads lead to less favorable attitudes toward both the ad and the advertised brand, as well as lower sharing intentions. Tellis et al. (2019) showed that this effect occurs because positively valenced content elicits positive emotions, such as love, pride, and joy, whereas negatively valenced content elicits negative emotions, such as fear, sadness, and shame. Therefore, we expect that the more negative the content of a meme is perceived to be, the less likely consumers are to share it. Hence, we predict the following:

• H_2a: The perceived negativity of a meme has a negative effect on meme virality.

Furthermore, we expect that the perceived negativity of memes about Airbnb negatively impacts the company’s warmth and competence images. We base these expectations on the well-documented negativity bias, which suggests that negative information is more diagnostic in impression formation (Rozin & Royzman, 2001). This proposition is supported by the meta-analysis by Purnawirawan et al. (2015) who provide evidence that the more negative an online review is, the more useful it is perceived to be. Moreover, several prior studies, also conducted in tourism contexts, show that (non-memetic) negative UGC has a negative effect on brand image (Bakri et al., 2020; Barreda & Bilgihan, 2013; Gensler et al., 2015). Applying the reasoning and findings of these studies to the context of memes, we predict that when Airbnb-related memes are perceived as negative, consumers find such memes more informative and lower their perceptions of Airbnb’s warmth and competence:

• H_2b: The perceived negativity of a meme has a negative effect on warmth image.

• H_2c: The perceived negativity of a meme has a negative effect on competence image.

Next, we hypothesize that meme polysemy, which refers to the existence of at least two distinct interpretations of the same meme, negatively affects meme virality. We base this on Shifman’s (2013) assertion that clear and simple memes are more likely to go viral. Our prediction can be explained from both synchronic and diachronic polysemy perspectives. Synchronic polysemy refers to alternative interpretations of a message “across two or more audiences at a certain point in time” (Puntoni et al., 2010, p. 52). If different groups of users interpret a meme differently, it is unlikely that each group will be equally inclined to share its inferred meaning. This implies that a meme that is uniformly interpreted by all users is (ceteris paribus) more likely to go viral. Diachronic polysemy refers to the multiplicity of inferred meanings of a message by a single user (Puntoni et al., 2010). When a user infers more than one meaning from a meme, they are also less likely to share that meme because their motive for doing so is less apparent (Jensen et al., 2020). This rationale is in line with Ge and Gretzel (2018), who observe that a message’s complexity negatively affects its virality. Therefore, we hypothesize:

• H_3a: The polysemy of a meme has a negative effect on meme virality.

We also expect that meme polysemy will have negative effects on a brand’s warmth and competence. Polysemic memes are more difficult to comprehend, negatively affecting brand associations and brand image (Shen et al., 2023; Teng et al., 2022). Additionally, findings from literature on polysemic marketing communications show that consumers perceive polysemic messages as more complex, ambiguous, and confusing (Dimofte & Yalch, 2020; Shahriari et al., 2023; Zúñiga et al., 2016). Since polysemic messages are more ambiguous (Puntoni et al., 2010), and ambiguous brand-related information leads to increased attitudinal ambivalence (Siddiqi et al., 2020), ambiguity negatively affects brand image (Grimm & Wagner, 2021). Accordingly, we predict the following:

• H_3b: The polysemy of a meme has a negative effect on warmth image.

• H_3c: The polysemy of a meme has a negative effect on competence image.

Intertextuality is the last meme characteristic for which we hypothesize the effect on meme virality. Various authors argue that shared references contribute to virality (Jenkins et al., 2013) and intertextuality facilitates the dissemination of memes (Laineste & Voolaid, 2016), which explains why successful memes are rich in intertextuality (Lankshear & Knobel, 2019). When a meme references popular culture and/or shares common features with an already viral meme(s), it invokes culturally shared knowledge, which makes it easier to comprehend (Wiggins, 2019). This rationale is in line with the findings of Feng and Wignell (2011), who report that the intertextuality of advertisements contributes to consumer engagement. Therefore, we propose:

• H_4a: The intertextuality of a meme has a positive effect on meme virality.

We also expect meme intertextuality to improve the warmth and competence image of a brand. We base this expectation on McCracken’s (1986) model of the structure and movement of the cultural meaning of consumer goods, which suggests that cultural meanings move from a culturally constituted world to consumer goods. When a brand-related meme includes cultural elements (e.g., image, quote, symbol, person), consumers will be more likely to ascribe the underlying cultural meaning to the brand. This proposition is supported by at least two streams of literature. The brand trait transference literature shows that consumers often attribute the traits possessed by celebrities to the brands with which they are associated (Bergkvist, 2017). In the consumer stereotyping literature, Gidaković et al. (2021) show that consumers develop perceptions of a brand’s warmth and competence by ascribing to it social (i.e., country stereotypes) and cultural (i.e., brand user stereotypes) meanings. Following these findings, as well as the logic of the model of the structure and movement of the cultural meaning of consumer goods (McCracken, 1986), we predict:

• H_4b: The intertextuality of a meme has a positive effect on warmth image.

• H_4c: The intertextuality of a meme has a positive effect on competence image.

Methodology

Research Design and Data Collection

To test the conceptual model (see Figure 2), we designed a multilevel study with the participants (Level-1) nested in memes (Level-2). We opted for this research design as it offers several advantages (c.f. Chaudhuri & Holbrook, 2001). First, it enables us to generate a dataset that includes a large sample of actual memes that were evaluated by a large sample of participants, which enhances the study’s ecological validity (van Heerde et al., 2021). Second, in comparison to an experimental design (Viglia & Dolnicar, 2020), our design enables testing the effects of several meme characteristics on brand doppelgänger image dimensions and meme virality while still controlling for individual differences among respondents and contextual differences among memes. Lastly, as described in more detail below, our research design accommodates a split-sample approach whereby each participant provides scores for the independent and dependent variables in relation to a different meme. This alleviates common method bias concerns (Min et al., 2016) and provides stronger evidence of any causal effects (Antonakis et al., 2014).

To obtain a representative sample of Airbnb-related memes, we followed methodological recommendations and prior practice regarding meme sampling (Laineste & Voolaid, 2016; Shifman, 2007), combining top-down and bottom-up approaches. For the former, we identified the most popular and relevant meme websites based on the Google page rank (e.g., knowyourmeme.com, 9gag.com, and boredpanda.com) and performed searches using the keyword “Airbnb” on each website. For the bottom-up approach, we performed a Google (image) search using the keyword “Airbnb” combined with “meme” or “humor,” which resulted in a preliminary sample of 201 unique Airbnb-related memes. Two authors then coded the content of each meme as positive, neutral, or negative (Cohen’s κ = .89). We excluded 38 positive memes from the sample as they did not correspond with the definition of doppelgänger brand image as a “family of disparaging images” (Thompson et al., 2006, p. 50). From the remaining sample of 163 negative and neutral memes, we randomly selected 121 memes, which we deemed to be an appropriate sample size at Level-2 of our hierarchical linear models (HLMs; Maas & Hox, 2005).

Next, we developed a four-part survey questionnaire that aligned with the proposed research model. The introductory part included informed consent for participants and a definition of memes, which ensured a common understanding of the phenomenon. In the second part, each participant was randomly assigned one meme from our sample of 121 memes and asked to evaluate it regarding the dependent variables (i.e., meme virality and brand image). In the third part, each participant was randomly assigned a different meme from our meme sample and asked to evaluate it regarding the independent variables. Each meme was evaluated by at least 27 participants for the dependent variables (the mode was 31 participants) and at least 25 participants for the independent variables (the mode was 30 participants). In the final part of the questionnaire, the participants answered questions related to the control and demographic variables. After pretesting the questionnaire on a small sample of participants (n = 50), we used Prolific Academic to recruit 3,664 participants from the United States (M_age = 38.45; SD_age = 13.46). Data collection took place in September of 2022. A total of 58.8% of participants reported having previously shared a meme (on any topic), while 16.6% had experience creating a meme. In addition, 56.5% of the participants had used Airbnb as a guest and 2.7% as a host.

The resulting dataset of memes and the participants’ survey responses can be accessed via the OSF platform (bit.ly/3N4pdrK). In addition to the previously described split-sample approach, we implemented procedural remedies to ensure data quality and minimize common method bias. For instance, we included a directed query attention check (Abbey & Meloy, 2017), which resulted in the exclusion of 40 inattentive participants. Moreover, we ensured participant anonymity and used brief scales that included a combination of different rating formats (MacKenzie & Podsakoff, 2012).

Operationalization and Modeling Approach

We used a combination of previously validated measurement scales and our own coding of the meme characteristics to operationalize the constructs in our conceptual model (Figure 2). We employed Likert-type scales from the literature to measure all three dependent variables (i.e., meme virality, warmth, and competence brand image). Furthermore, we relied on multi-item scales to operationalize three (of the four) independent variables with hypothesized effects (i.e., perceived meme negativity, humor, and polysemy). Panel A in Table 1 summarizes the items and scale reliabilities. In addition to the participants’ demographic characteristics (age, gender, income, and educational level), we included single-item measures of their experiences with memes and Airbnb. As explained subsequently, we included these characteristics as participant-level control variables in our HLM models.

Table 1.

Operationalization of the Variables.

Panel A: survey questionnaire, scale/item	Factor loading
Meme virality (Pennycook et al., 2018):
Would you consider sharing this meme online, for example, through Facebook or Twitter?	—
Warmth image (van Prooijen & Bartels, 2019); α = .97:
The above meme portrays Airbnb as likeable.	0.94
The above meme portrays Airbnb as friendly.	0.97
The above meme portrays Airbnb as warm.	0.95
Competence image (van Prooijen & Bartels, 2019); α = .96:
The above meme portrays Airbnb as intelligent.	0.94
The above meme portrays Airbnb as skilled.	0.95
The above meme portrays Airbnb as competent.	0.94
Perceived meme negativity (Thomas & Fowler, 2021); α = .89:
This meme is mean-spirited.	0.84
This meme is negative.	0.87
This meme is ridiculing.	0.86
Perceived meme humor (Thomas & Fowler, 2021); α = .77:
This meme is humorous.	0.66
This meme is teasing.	0.88
This meme is sarcastic.	0.67
Meme polysemy (Arora et al., 2015; Puntoni et al., 2010); α = .80:
The meaning of this meme is ambiguous/unambiguous.	0.86
The meaning of this meme is unclear/clear.	0.83
The meaning of this meme is multifaceted/uniform.	0.60
Meme creator
Have you ever created a meme (on any topic)?	—
Meme sharer
Have you ever shared a meme (on any topic)?	—
Airbnb host
Have you ever used Airbnb as a host?	—
Airbnb guest
Have you ever used Airbnb as a guest?	—
Panel B: coding manual with interrater reliabilities, Variable/category definition	% of meme sample
Meme intertextuality (Feng & Wignell, 2011; Laineste & Voolaid, 2016); Cohen’s κ = .88:
Intertextual = the meme is referring to existing cultural phenomenon or includes some pre-existing content (text, picture, or symbol).	52.1%
Not intertextual = the meme does not include any borrowed content and does not refer to existing cultural phenomena.	47.9%
Meme target (own content analysis); Cohen’s κ = .83:
Platform = the meme refers to Airbnb as a platform and phenomenon that includes guests and hosts as categories/groups of actors.Host = the meme refers to an identifiable person who has a listing on Airbnb.	47.9%
Guest = the meme refers to an identifiable person who is searching for or renting a place.	37.2%
Multiple targets = more than one specific actor.	10.7%
Meme form (Shifman, 2013); Cohen’s κ = .91:	4.1%
Picture = only images that speak for themselves, without any verbal elements or text.	3.3%
Text = 80% of a meme or more is text but may also include small pictures.	10.7%
Captioned image = image or symbol that only makes sense in combination with text.	86%
Meme theme (own content analysis); Cohen’s κ = .78:
Playful fun = co-creation through suggestive play or verbal and/or visual humor.	32.2%
Horror = showing horrific and spooky appearances of places, including rental conditions.	21.5%
Ridiculous = misleading, weird, and fraudulent policies, including misleading presentation of places or prices and/or inadequate services.	23.1%
Offensive = discrimination, racism, disrespect, or unacceptable conduct or communication.	7.4%
New economy = representative symbolism of sharing economy, including socially irresponsible ideology or exploitative capitalism.	15.7%

Note. — = factor loading could not be computed because a variable was measured with a single item. All the standardized factor loadings are significant at p < .001.

In addition to the variables measured through our survey questionnaire, we coded several meme characteristics. Following the approach of previous meme content analysis studies (Brubaker et al., 2018; Kolar & Walanchalee, 2020), we developed coding categories for the variables of meme intertextuality, meme content themes, meme form, and meme target. Two authors then independently coded each of the 121 memes in our sample using these four variables. Panel B of Table 1 summarizes the definitions of the coding categories and indicates that the interrater reliability was appropriate (Cohen’s κ ranging between .78 and .91).

To assess the validity of our survey measures, we used the scores from the entire sample of 3,664 participants and performed a confirmatory factor analysis (CFA) in Amos (version 28). The results of the CFA indicated that the model fit the data well (χ²_{(df = 90)} = 1,654.07; CFI = 0.96; TLI = 0.95; RMSEA = 0.07; SRMR = 0.06). Moreover, Table 2 shows that all the measurement scales were reliable and conformed to the standards for convergent and discriminant validity (Fornell & Larcker, 1981). As we tested our hypotheses using a hierarchical linear regression framework (Raudenbush & Bryk, 2002), which did not support factor analysis, we computed the factor score weights for all the constructs that were operationalized with multi-item scales. We then calculated the mean factor score weights for the perceived meme negativity, perceived meme humor, and meme polysemy for each of the 121 memes in our sample and used these variables as inputs for the HLMs. For the multi-categorical variables (i.e., meme-level controls), we created dummies and used the following categories as baselines: multiple targets for meme target, captioned image for meme form, and new economy for meme content theme. Table 3 presents descriptive statistics and correlation coefficients for all the meme-level (Level-2) variables.

Table 2.

Individual Respondent-Level Correlations.

Construct	CR	AVE	1	2	3	4	5	6	7	8	9	10	11	12	13	14
Meme virality (1)	—	—	—
Warmth image (2)	0.97	0.91	0.24***	0.95
Competence image (3)	0.96	0.89	0.26***	0.73***	0.94
Perceived meme negativity (4)	0.89	0.73	−0.01	−0.10***	−0.07***	0.86
Perceived meme humor (5)	0.79	0.55	0.12***	0.05*	0.06***	0.10***	0.74
Meme polysemy (6)	0.81	0.60	0.01	0.00	0.01	−0.05*	0.15***	0.77
Meme creator (7)	—	—	−0.05**	0.01	0.01	0.06***	−0.03	−0.06***	—
Meme sharer (8)	—	—	−0.07***	0.02	0.01	0.07***	−0.05**	−0.04*	0.33***	—
Airbnb host (9)	—	—	0.07***	0.05***	0.06***	−0.01	0.04*	−0.03 #	−0.04*	−0.01	—
Airbnb guest (10)	—	—	0.03 #	0.00	0.01	−0.02	0.04*	−0.02	−0.08***	−0.10***	0.10***	—
Age (11)	—	—	0.00	0.00	−0.02	0.15***	−0.03 #	−0.07***	0.30***	0.28***	−0.02	−0.19***	—
Gender (12)	—	—	−0.05*	−0.03 #	−0.02	0.01	−0.04*	−0.04*	0.08***	−0.06***	−0.02	−0.01	−.02	—
Education (13)	—	—	−0.04*	−0.05**	−0.04*	0.09***	0.03 #	−0.05*	0.06***	0.09***	0.05**	0.2***	.11***	−.06***	—
Income (14)	—	—	0.03 #	−0.02	0.00	0.06***	0.02	−0.06**	0.01	0.00	0.04*	0.09***	.05**	−.06***	.18***	—
Variable mean	—	—	1.81	3.17	3.07	3.94	4.41	4.53	1.59	0.41	0.83	0.56	38.45	.47	2.94	3.19
Variable standard deviation	—	—	1.16	1.57	1.41	1.36	1.11	1.49	.37	0.49	0.15	49	13.46	.57	.78	1.78

Note. CR = composite reliability; AVE = average variance extracted. — = a statistic cannot be computed as the variable was measured with a single item. Square root of the AVE in bold on the diagonal. N = 3,664.

Statistical significance levels: ^# p < .10. *p < .05. **p < .01. ***p < .001.

Table 3.

Meme-level correlations.

Construct	1	2	3	4	5	6	7	8	9	10	11	12	13
Perceived meme negativity (1)	1
Perceived meme humor (2)	0.08	1
Meme polysemy (3)	0.09	0.09	1
Meme intertextuality (4)	0.03	−0.04	0.00	1
Meme target—platform dummy (5)	0.06	−0.09	−0.11	0.59**	1
Meme target—guest dummy (6)	−0.22*	−0.08	0.19*	−0.15	−0.33**	1
Meme target—host dummy (7)	0.08	0.16	−0.01	−0.46**	−0.74**	−0.27**	1
Meme form—picture dummy (8)	−0.15	−0.05	0.02	−0.10	−0.09	−0.06	0.05	1
Meme form—text dummy (9)	0.10	−0.06	0.07	−0.31**	0.15	−0.03	−0.16	−0.06	1
Meme theme—playful fun dummy (10)	−0.15	−0.07	−0.02	0.10	−0.06	0.16	−0.02	−0.03	−0.07	1
Meme theme—horror dummy (11)	0.15	0.07	−0.20*	−0.30*	−0.22*	−0.05	0.31**	−0.10	−0.18*	−0.36**	1
Meme theme—ridiculous dummy (12)	−0.15	0.07	−0.02	0.06	0.02	−0.13	0.06	0.12	−0.00	−0.38**	−0.29**	1
Meme theme—offensiveness dummy (13)	0.02	−0.17	0.15	−0.04	−0.15	0.21*	−0.09	0.12	0.00	−0.20*	−0.15	−0.16	1
Mean	0.00	0.00	0.00	0.52	0.48	0.11	0.37	0.03	0.11	0.32	0.22	0.23	0.07
Standard deviation	0.24	0.20	0.28	0.50	0.50	0.31	0.49	0.18	0.31	0.47	0.41	0.42	0.26

Note. N = 121.

Statistical significance levels: *p < .05. **p < .01.

We used HLM software (version 8.0) to estimate three mixed models that differed only with respect to the dependent variable used (i.e., meme virality/warmth image/competence image). We specified the following equation for each model:

$$\begin{array}{c}\multicolumn{1}{c}{\mathrm{Meme}\mathrm{viralit}{\mathrm{y}}_{\mathrm{ij}}/\mathrm{Warmth}\mathrm{imag}{\mathrm{e}}_{\mathrm{ij}}/\mathrm{Competence}\mathrm{imag}{\mathrm{e}}_{\mathrm{ij}}={\mathrm{\gamma }}_{00}+{\gamma }_{01}\times \left(\mathbf{Perceived}\mathbf{meme}\mathbf{humo}{\mathbf{r}}_j\right)}\\ \multicolumn{1}{c}{+{\gamma }_{02}\times \left(\mathbf{Perceived}\mathbf{meme}\mathbf{negativit}{\mathbf{y}}_j\right)+{\gamma }_{03}\times \left(\mathbf{Meme}\mathbf{polysem}{\mathbf{y}}_j\right)+{\gamma }_{04}\times \left(\mathbf{Meme}\mathbf{intertextualit}{\mathbf{y}}_j\right)}\\ \multicolumn{1}{c}{+{\gamma }_{05}\times (\mathbf{Meme}\mathbf{target}-\mathbf{platform}\mathbf{dumm}{\mathbf{y}}_j)+{\gamma }_{06}\times (\mathbf{Meme}\mathbf{target}-\mathbf{guest}\mathbf{dumm}{\mathbf{y}}_j)}\\ \multicolumn{1}{c}{+{\gamma }_{07}\times (\mathbf{Meme}\mathbf{target}-\mathbf{host}\mathbf{dumm}{\mathbf{y}}_j)+{\gamma }_{08}\times (\mathbf{Meme}\mathbf{form}-\mathbf{picture}\mathbf{dumm}{\mathbf{y}}_j)}\\ \multicolumn{1}{c}{+{\gamma }_{09}\times (\mathbf{Meme}\mathbf{form}-\mathbf{text}\mathbf{dumm}{\mathbf{y}}_j)+{\gamma }_{010}\times (\mathbf{Meme}\mathbf{theme}-\mathbf{playful}\mathbf{fun}\mathbf{dumm}{\mathbf{y}}_j)}\\ \multicolumn{1}{c}{+{\gamma }_{011}\times (\mathbf{Meme}\mathbf{theme}-\mathbf{horror}\mathbf{dumm}{\mathbf{y}}_j)+{\gamma }_{012}\times (\mathbf{Meme}\mathbf{theme}-\mathbf{ridiculous}\mathbf{dumm}{\mathbf{y}}_j)}\\ \multicolumn{1}{c}{+{\gamma }_{012}\times (\mathbf{Meme}\mathbf{theme}-\mathbf{offensiveness}\mathbf{dumm}{\mathbf{y}}_j)+{\gamma }_{10}\times \left(\mathrm{Meme}\mathrm{share}{\mathrm{r}}_{\mathrm{ij}}\right)+{\gamma }_{20}\times \left(\mathrm{Meme}\mathrm{creato}{\mathrm{r}}_{\mathrm{ij}}\right)}\\ \multicolumn{1}{c}{+{\gamma }_{30}\times \left(\mathrm{Airbnb}\mathrm{gues}{\mathrm{t}}_{\mathrm{ij}}\right)+{\gamma }_{40}\times \left(\mathrm{Airbnb}\mathrm{hos}{\mathrm{t}}_{\mathrm{ij}}\right)+{\gamma }_{50}\times \left(\mathrm{Ag}{\mathrm{e}}_{\mathrm{ij}}\right)+{\gamma }_{60}\times \left(\mathrm{Gende}{\mathrm{r}}_{\mathrm{ij}}\right)+{\gamma }_{70}}\\ \multicolumn{1}{c}{\times \left(\mathrm{Educatio}{\mathrm{n}}_{\mathrm{ij}}\right)+{\gamma }_{80}\times \left(\mathrm{Incom}{\mathrm{e}}_{\mathrm{ij}}\right)+{\mu }_{0\mathrm{j}}+{\mathrm{e}}_{\mathrm{ij}},}\end{array}$$ (1)

where i = 1, 2, 3 …I participants, and j = 1, 2, 3 …J memes. The Level-2 predictors shown in bold were centered on the grand mean to make the intercepts γ₀₀“more interpretable” (Hofmann, 1997, p. 738). Multicollinearity is a major threat to the validity of multilevel modeling, but HLM software does not offer a way to assess it (Raudenbush & Bryk, 2002). Therefore, we inspected the correlation coefficients between the independent variables at Level-1 (Table 2) and Level-2 (Table 3). The correlations among Level-1 predictors were relatively low (largest r = .15), but some correlations among the Level-2 predictors were higher than .5. Thus, we computed the variance inflation factors (VIFs). Only the VIFs of the dummies representing the meme target control variable were larger than 3 (ranging between 3.6 and 8.4), but these were still below the recommended VIF threshold of 10 (O’Brien, 2007). Hence, we decided to retain this control variable in all three models.

Results

Table 4 presents the results for all three HLMs. We first calculated the intraclass correlation coefficients, which ranged between .07 and .24 and were thus above the recommended .05 threshold that warrants multilevel modeling (Heck & Thomas, 2015). We then proceeded with the inspection of the parameter estimates. In the first model, we observe a significant positive effect of perceived humor (γ = .389, p < .05) and a significant negative effect of meme polysemy (γ = −.286, p < .05) on meme virality, which supports H_1a and H_3a, respectively. The effects of perceived meme negativity and meme intertextuality are not statistically significant, which lead us to reject H_2a and H_4a. We also find significant effects of the control variables at both the participant and meme levels. ² The results show that participants who have higher disposable income (γ = .021, p < .1) or experience with meme sharing (γ = .190, p < .001), Airbnb hosting (γ = .461, p < .001), and/or renting (γ = .072, p < .1) are more likely to share Airbnb-related memes. However, female participants are less likely to share Airbnb memes than male participants (γ = −.066, p < .1), as are participants with higher levels of education (γ = −.048, p < .05). Furthermore, we find that Airbnb-related memes are less likely to be shared when they target the platform (γ = −.353, p < .01) or a specific host (γ = −.309, p < .05) rather than targeting a specific guest or multiple actors. We also find Airbnb-related memes are more likely to be shared when they include only pictures (γ = .477, p < .05) than when they feature text or captioned images. Lastly, memes with offensive content (γ = −.343, p < .05) are less likely to be shared than memes with other content themes.

Table 4.

Results of the Hierarchical Linear Models.

Predictor (hypothesis)	Dependent variable
	Meme virality	Warmth image	Competence image
Intercept	1.817*** (58.769)	0.002 (0.024)	0.003 (0.045)
Individual participant-level (Level-1) control variables
Meme creator	0.059 (−1.297)	0.016 (0.316)	0.041 (0.923)
Meme sharer	0.190*** (−3.427)	0.094 (1.536)	0.057 (0.971)
Airbnb host	0.461** (2.787)	0.441* (2.413)	0.506*** (3.303)
Airbnb guest	0.072 # (1.863)	0.036 (0.877)	0.011 (0.312)
Age	0.003* (2.035)	0.002 (0.754)	−0.001 (−0.730)
Gender	−0.066 # (−1.946)	−0.092* (−2.260)	−0.069 # (−1.823)
Education	−0.048* (−2.230)	−0.064* (−2.409)	−0.049 # (−1.889)
Income	0.021 # (1.844)	−0.001 (−0.100)	0.007 (0.056)
Meme-level (Level-2) control variables
Meme target—platform dummy	−0.353** (−2.459)	−0.565 (1.458)	−0.532 (−1.442)
Meme target—guest dummy	−151 (−0.961)	−0.242 (−0.600)	−0.297 (−0.743)
Meme target—host dummy	−0.309* (−2.369)	−0.400 (−1.103)	−0.389 (−0.356)
Meme form—picture dummy	0.477* (2.209)	0.055 (0.209)	−0.180 (−0.554)
Meme form—text dummy	−0.058 (−0.411)	0.263 (0.991)	0.380 (1.559)
Meme theme—playful fun dummy	−0.148 (−1.576)	0.487* (2.275)	−0.031 (−0.174)
Meme theme—horror dummy	−0.182 (−1.534)	−0.457 # (−1.700)	−0.599** (−2.724)
Meme theme—ridiculous dummy	−0.012 (−0.117)	−0.397 # (−1.674)	−509** (−2.804)
Meme theme—offensiveness dummy	−0.343* (−2.448)	−0.358 (−0.897)	−0.265 (−0.787)
Meme-level (Level-2) hypothesized effects
Perceived meme humor	0.389* (2.530)	0.051 (1.372)	0.471 (1.341)
Perceived meme negativity	0.081 (.545)	−1.058*** (−3.785)	−0.768** (−3.100)
Meme polysemy	−0.286* (−2.431)	−0.075 (−0.301)	−0.147 (−0.708)
Meme intertextuality	−0.133 (−1.363)	−0.111 (−0.580)	0.057 (0.354)
Deviance	11,297.300	12,661.842	12,245.534
Pseudo R2	.040	.119	.069
Intraclass correlation coefficient	.070	.240	.194

Note. T-ratios in parentheses.

Statistical significance levels: ^# p < .10. *p < .05. **p < .01. ***p < .001.

In the second HLM, we find a significant negative effect (γ = −1.058, p < .001) of perceived meme negativity on warmth image, which supports H_2b. However, the effects of perceived meme humor, meme polysemy, and meme intertextuality are not statistically significant, leading us to reject H_1b, H_3b, and H_3c. We again observe significant effects of certain control variables. More specifically, we find that Airbnb-related memes have stronger effects on warmth image when participants have experience hosting through Airbnb (γ = .441, p < .05). Conversely, the effect of memes on warmth image is weaker for female participants than male participants (γ = −.092, p < .05) and for participants with higher levels of education (γ = −.064, p < .05). We also find that the effect of memes on warmth image is positive when the content theme of a meme is playful fun (γ = .487, p < .05) and negative when the content theme is horror (γ = −.457, p < .1) or ridiculous (γ = −.397, p < .1).

In the third HLM, we find support for H_2c as the effect of perceived meme negativity on competence image is significant and negative (γ = −.768, p < .01). The results in Table 4 lead us to reject H_1c, H_3c, and H_4c as the effects of perceived meme humor, meme polysemy, and meme intertextuality on competence image are not statistically significant. We observe a similar pattern as with warmth image when we turn to the control variables’ effects on competence image. Specifically, we find that memes have a stronger effect on Airbnb’s competence image when the participants have experience hosting through Airbnb (γ = .506, p < .001). However, the effect of memes on warmth image is weaker for female participants than male participants (γ = −.069, p < .1), and for participants with higher levels of education (γ = −.049, p < .1). We again find that the effect of memes on warmth image is negative when the meme content theme is horror (γ = −.599, p < .01) or ridiculous (γ = −.509, p < .01).

Discussion

In this paper, we set out to examine how diverse conceptual and content-related characteristics of memes, as well as the individual characteristics of meme users, influence meme virality and brand imagery in the case of Airbnb. Theoretically, we based this investigation on the integration of the doppelgänger brand concept and meme virality framework (Malodia et al., 2022) into MacInnis and Jaworski’s (1989) information processing framework. The results of our hierarchical linear modeling are presented in Table 4, and the corresponding summary of our hypotheses testing appears in Table 5. In sum, the results suggest that the virality of disparaging memes is driven mostly by their humorous style and clear meaning, coupled with their pictorial and inoffensive content and “active” user status (i.e., Airbnb hosts and meme creators/sharers). Perceived brand image is, however, affected mostly by the stance (negativity) of memes, for which additional content themes are important (i.e., playful fun, horror, ridiculous), and “active” user status again affects brand perception.

Table 5.

Overview of the Hypothesis Testing.

Hypothesis	Empirically supported
H1a: The perceived humor of a meme has a positive effect on meme virality.	✓
H1b: The perceived humor of a meme has a positive effect on warmth image.	✗
H1c: The perceived humor of a meme has a positive effect on competence image.	✗
H2a: The perceived negativity of a meme has a negative effect on meme virality.	✗
H2b: The perceived negativity of a meme has a negative effect on warmth image.	✓
H2a: The perceived negativity of a meme has a negative effect on competence image.	✓
H3a: The polysemy of a meme has a negative effect on meme virality.	✓
H3b: The polysemy of a meme has a negative effect on warmth image.	✗
H3c: The polysemy of a meme has a negative effect on competence image.	✗
H4a: The intertextuality of a meme has a positive effect on meme virality.	✗
H4b: The intertextuality of a meme has a positive effect on warmth image.	✗
H4c: The intertextuality of a meme has a positive effect on competence image.	✗

Theoretical Implications

The results obtained have several theoretical implications, contributing to the literature at the intersection of memes, UGC, and viral tourism branding. First, this study and its findings provide more general contributions to the tourism field because they address some of the important co-branding challenges identified by Munar (2011) and Buhalis and Park (2021). More specifically, in the “era of brand love” (see Aro et al., 2018; Bigne et al., 2020; Foroudi et al., 2023; Mody & Hanks, 2019), it is of critical importance to also understand the darker, disparaging, and negative side of co-branding illuminated by this study, which reveals that Airbnb is depicted in a range of different themes, including playful fun, horror, ridicule, and offensiveness. As such, the doppelgänger construct complements other, more aggressive, and more extreme concepts addressed recently, such as brand hijacking (Siano et al., 2022), brand hate (Kucuk, 2019b), brand attack (Rauschnabel et al., 2016), brand monsters (Freund & Jacobi, 2013), and brand hitlerization (Kucuk, 2020).

Furthermore, the literature on memes and doppelgängers emphasizes the importance of the social (cultural, communal) component of both, which renders them relevant to understanding branding in the sharing economy, especially for brands, such as Airbnb, that aspire to foster a shared identity. In such (platform, customer-to-customer) contexts, branding becomes a community affair and brands cannot control the narrative. This is why the symbolic, emotional, and bonding functions of brands are becoming increasingly important, with brands aspiring to act and be perceived as in-group members (Halkias, 2022).

Looking into the findings in more detail, the examination of the hypothesized antecedents of meme virality confirms the positive influence of humor and the negative influence of polysemy on meme virality. Memes that involve humor and that are less open to interpretation will be shared more. These findings enable us to highlight the role of humor and polysemy in meme virality, as emphasized in prior studies (e.g., Ge & Gretzel, 2018; Malodia et al., 2022). Contrary to our expectations, neither intertextuality nor perceived negativity had an effect on meme virality. The participants in our study did not express higher sharing intentions for intertextual memes that referenced popular culture and/or involved pre-existing content. This implies that participants may appreciate the individual creativity of meme creators as much as any reference to existing cultural artifacts, meaning that they are not inclined to share one type of meme more than others. Moreover, the participants’ intention to share memes was the same regardless of how negative the memes were, as other factors held more weight. This is surprising, given that the literature on memes suggests that positive online content is more viral than negative online content (Shifman, 2014; Berger & Milkman, 2012). A possible explanation can be that the study by Berger and Milkman (2012) explored different types of content (e.g., advertisements, videos, and news articles), which did not account for the humor of the content and included a more limited number of user-related (creator and/or sharer) variables.

We also contribute to the literature on brand information processing that relies on MacInnis and Jaworski’s framework (1989) by demonstrating the relevance of extending its original antecedents with meme characteristics. As expected in our hypotheses, we confirm the influence of meme negativity on both the warmth and competence dimensions of brand image. The more negatively perceived the portrayal of the brand in a meme is, the more negative the image perceptions of the brand’s warmth and competence are (and vice versa). These results echo those of Brandt et al. (2010), who confirm the effects of negativity on brand image with regard to the doppelgänger concept.

However, humor, polysemy, and intertextuality have no effect on the two dimensions of brand image. This finding does not align with the findings of the meta-analysis by Eisend (2009), who reports positive effects of humor in advertising on brand attitudes. Thus, memetic humor may be specific and not beneficial or harmful for branding purposes per se. The lack of support for the humor-related hypotheses in this study may exist for various reasons. Additional nuances of humor that were not controlled for can also be relevant factors of virality, such as humor type and style. For instance, parody is one of the most characteristic and viral styles of contemporary UGC (Vanden Bergh et al., 2011), and when used strategically, it may represent a real threat to the brand (Jean, 2011). Different styles of memetic humor (not controlled for in our study) were, however, found to affect virality differently (Taecharungroj & Nueangjamnong, 2015). In addition, there are differences in the sources and nature of advertising and memes, with the source of advertising being the brand, and the source of memes being the individual meme creator. References to the cultural context are essential for memes (Shifman, 2013, 2019; Tanni, 2020), while the commercial context of advertising differs notably from the non-commercial (grassroots) context in which memes flourish. As a result, in the eyes of the meme recipients, the brand’s imagery may be detached from its doppelgänger memes. This might also explain why intertextuality and polysemy are not significant predictors, as the link with a brand’s characteristics may be too weak to influence its imagery. Another reason for the lack of support for intertextuality may be the perception of intertextuality, which may be balanced out by creativity. As previous studies of other media formats have shown, creativity is an important determinant of consumers’ responses to brand-related communications (Rosengren et al., 2020). This suggests that creativity might be a potential factor that can explain the relationship between intertextuality and brand image when originally created and imitative (referring to pop-cultural resources) memes are compared.

In terms of the meme-characteristic controls, we are able to provide additional support for the role of stimulus-related characteristics and behavioral outcomes (i.e., sharing) as extensions of the information processing model (MacInnis & Jaworski, 1989). Memes that are less offensive and that include a picture will be shared more often. The negative influence of offensiveness mirrors the results of previous studies in the advertising field, with offensive ads having a negative impact on consumer responses (Chan et al., 2007; Prendergast & Hwa, 2003). The findings suggest that only some negative themes affect virality. More nuanced distinctions among content in terms of the themes therefore prove to be justified, as the offensive theme is the only negative theme that has a significant (negative) impact on virality, while the impacts of horror and ridicule are not significant.

Additionally, when the differences between picture- and text-based ads were considered in previous studies, it was found that they may depend on the stage of the consumer decision-making process. By finding support for the role of pictures in meme virality, our results conform to the finding that pictures may be more successful in generating initial arousal and interest (Lewis et al., 2013). This somewhat contradicts the findings concerning the use of concrete versus abstract metaphors in ads, where the use of concrete metaphors is more successful (Morgan & Reichert, 1999).

We are also able to demonstrate that the meme themes influence brand warmth and competence, with playfulness having a positive effect on brand warmth and both ridiculousness and offensiveness having a negative effect on brand competence. These findings are in line with those of Kristal et al. (2018), who find that non-collaborative co-creation in the form of a more aggressive and straightforward “attack” results in a stronger dilution of brand equity than brand parody and play.

As for the control variables pertaining to the participants’ individual characteristics, we identify several factors that are more relevant in driving meme virality and brand imagery. For example, Airbnb hosts are more likely to share content. For the hosts, memes also have stronger effects on brand warmth and competence than for the non-hosts. Airbnb hosts exhibit a stronger self-brand connection, belong to the community, and as a result, are also more susceptible to memes that involve Airbnb. Unsurprisingly, memes are also more likely to be shared by people who have shared memes in the past. Additional predictors of both meme virality and brand image are gender and education. Here, males and users with lower educational levels are more inclined to share memes virally and to perceive both brand dimensions more positively. Of further interest is the finding that age has no impact on either virality or perceived brand image, given that some studies found age to affect virality (Wong & Holyoak, 2021).

In sum, our findings suggest that viral disparagement should be “mild, clear, and vivid,” given that humorous and pictorial UGC is a more effective (viral) means of co-branding than offensive UGC, while the valence of the content is more relevant for positive brand perceptions. The findings also suggest that memetic and disparaging co-branding is not merely a reversed version of the positive branding done by advertising but depends on the distinctive characteristics of the UGC examined. Accordingly, the additional roles of humor in tourism and branding are revealed, thus advancing and complementing the existing understanding of the topic (Ge & Gretzel, 2018).

Practical Implications

The current study has several implications for practitioners who could help develop brands and UGC strategies in the tourism sector, including brand managers, content (meme) creators, and media managers. Some of these implications are tactical and stem directly from the present findings, while other implications are more general and strategic.

Regarding the tactical implications, it has been suggested that consumers tend to ignore and feel disconnected from campaigns that are tightly controlled by brands (Malodia et al., 2022), indicating that memes provide an opportunity to engage and more freely interact with a brand. In terms of brand image, sharing-economy brand managers who are interested in starting new promotional campaigns may want to encourage and reward UGC that does not include negativity. As claimed elsewhere, UGC provides evidence that the power asymmetry between consumers and organizations is decreasing, with more power accumulating in the hands of consumers. It is easier and more sensible for brand managers to choose meme templates that signal positivity and avoid ridicule and offensiveness, thereby supporting brands and their images.

The individual-related characteristics that act as control variables in our framework may help tourism marketing managers segment the audience and better understand who will respond to certain memes more favorably. By knowing key consumer characteristics, marketers can increase the acceptance of memes by various customer segments. More specifically, men, consumers with lower educational levels, and Airbnb hosts are the most susceptible to meme marketing campaigns and are thus expected to engage in such campaigns more often. However, for consumers at the other end of the spectrum (e.g., women, consumers with higher educational levels, and non-hosts), tourism marketers need to rely on other forms of meme marketing (Razzaq et al., 2023, 2024).

Meme creators who either respond to a UGC campaign or simply engage in meme creation in the hope of going viral have several mechanisms to make their content more popular. It is evident that humor plays a key role in engaging an audience (Ge & Gretzel, 2018) and spreading a meme virally, as does the picture format. Including visual representations in memes leads to more positive responses. Furthermore, memes need to be simple and clear to become popular. While polysemy is a common characteristic of memes, the traditional wisdom that clear and concise meanings lead to more effective communication still holds true in the era of UGC communications. On the other hand, creators need to avoid making offensive memes or directly targeting Airbnb or its hosts if they want their work to be shared.

As for the strategic implications of this study, while memes are a type of UGC that is especially difficult to control, they still play an important role in marketing efforts, as they are insightful, engaging, and can enable authentic representations of brands. This demonstrates their importance in applying a cultural model of branding (Csordás et al., 2017).

As such, brand managers need to develop and operationalize adequate metrics (indicators) of brand health and equity. Accordingly, the proposed scales that have been shown to provide valid measures of the main dependent constructs of brand equity (i.e., virality, warmth, and competence), are deemed useful as they evaluate universal dimensions of brand image and can be effectively applied for comparative (competitive) evaluations. Moreover, memes can be used as an alternative research tool that can help identify (alternative) meanings associated with the brand, and by identifying brand “avoiders” and doppelgängers, provide early warning signals in case a brand is beginning to lose its cultural resonance (Thompson et al., 2006). Subsequently, when a new brand story is developed, memetic responses can help attract audience engagement and test responses to the new story. The development and sharing of a brand story and related UGC content are possible and warranted throughout the entire process of a tourism experience, including the before, during, and after stages of interaction with the brand.

Limitations and Future Research

Some limitations of our study need to be addressed, providing potential opportunities for future research. First, we focused only on memes related to a single sharing economy platform, which may limit the generalizability of our findings. Testing the proposed conceptual model for other sharing economy platforms (e.g., Uber) or in other tourism contexts (e.g., destination brands) may provide new and different insights into the phenomenon. Second, we included a general sample from the United States in our study, not from a particular community. Future studies could thus focus on more specific communities, such as hosts and other relevant groups (see Munar, 2010), or on respondents from other cultural backgrounds. Testing our conceptual model in different countries could provide valuable insights into any cross-cultural differences among the drivers of meme virality and brand image. Third, while our research design enabled an empirical investigation of the effects of various meme- and participant-related characteristics on meme virality and brand image, it is still considered a cross-sectional research design. Thus, future studies should employ experimental designs to provide stronger causal evidence of the significant drivers of meme virality and brand image identified in this paper. Moreover, future studies could sample memes from a single source that provides information regarding meme popularity/virality, which can be used as an “objective” dependent variable. As we focused on the doppelgänger brand image, capturing Airbnb’s image and its antecedents/consequences was beyond the scope of our study. Hence, future studies could also measure the perception of the tourist brand in question and/or examine the influence of brand stereotypes on the perception of UGC (memes). Alternatively, it would be also relevant to examine how doppelgänger image projected by UGC (memes) affects stereotypes about a particular brand, given that brand stereotyping is a dynamic social process (Gidaković et al., 2021; Halkias, 2022; Kervyn et al., 2021). Lastly, our study examined a range of meme- and participant-related constructs, although on a more general level. However, additional meme-, community-, or brand-related drivers could be included to capture individuals’ relationships with other users and brands (i.e., involvement, congruity) and help explain meme virality and perceived tourism brand image.

Finally, memes are conceptually an extremely complex type of UGC, where some authors point out up to 17 relevant conceptual characteristics (Cancelas-Ouviña, 2021), all potentially relevant for examination. Furthermore, numerous nuances of humor (over 40 different types, according to Buijzen & Valkenburg [2004]) also matter, which is why examining the effects of different styles and types of humor on virality and brand image is also advised in the future. Such complexity is certainly challenging for research, yet it reflects the reality that tourist brands face on social media today.