Keywords

1 Introduction

In a predominantly digital world, social media has become an essential part of people’s lives. As of October 2023, the global number of social media users was 4.95 billion, constituting 61.4% of the world’s population (Kepios 2024). The increased use of social media has prompted companies to adapt their customer communication, as social media platforms have a significant influence on users’ purchasing decisions (Schmidt/Taddicken (2017). Companies are increasingly using influencers, hereinafter referred to as human influencers (HIs), as an effective communication tool to reach and influence consumers (Ki/Kim 2019). Instagram in particular is the most popular social media channel for influencer marketing (Dencheva 2023).

In addition to the rapid growth of social media platforms and the success of HIs, virtual influencers (VIs) have started to emerge. A VI we define as an artificial representation of a human being. They can be divided into two broad categories, the anime virtual influencers (AVIs) and the human virtual influencers (HVIs). Most VIs used in influencer marketing are human virtual influencers and these also form the focus of this study. HVIs resemble real people and act on social media platforms such as HIs (Wibawa 2022). However, HVIs do not act autonomously, but are controlled by companies and are computer-generated, animated media personalities that have their own social media profile (Hofeditz et al. 2022). A team of developers designs the visual and verbal character of an HVI and its “life” story, considering authentic characteristics (Drenten/Brooks 2020). HVIs do not exist outside of the digital world and are legally required to communicate that they are a virtual entity (Park et al. 2021).

HVIs are particularly well known on image-based platforms such as Instagram, as they attract a lot of attention due to the human traits of their physical appearance, behaviour and personality, as well as through interaction with their followers (Moustakes et al. 2020). Examples of successful HVIs include Lu of Magalu, Lil Miquela and Shudu. They also have millions of followers and have already collaborated with various companies: Lu of Magalu, for example, had 6.4 million Instagram followers in 2023 and has collaborated with Samsung, McDonald’s, Vogue, Adidas and Lancome, among others. In that same year, Lil Miquela had 2.6 million Instagram followers and collaborations with Alexander McQueen, Gucci, Prada, Calvin Klein, Karl Lagerfeld, and more. Shudu, the world’s first black virtual influencer, had 241 thousand Instagram followers and has worked with luxury fashion brands Karl Lagerfeld and Louis Vuitton as well as with BMW (Storyclash 2023). It can therefore be assumed that HVIs also make a positive contribution to the advertised product with their appearance, their behaviour and the open communication of their virtual entity.

Compared to their real-life counterparts, HVIs have several advantages for companies: HVIs can be customised with their appearance and personality to fit the target group (Kádeková/Holienčinová 2018). Due to their immateriality, they can be used anywhere and at any time (Berrryman et al. 2021). HVIs also offer particular security, as they themselves cannot exhibit unplanned and inappropriate behaviour or have a negative past that could damage the brand (Drenten/Brooks 2020). In addition, companies are perceived as innovative and technically adept if they create or collaborate with an HVI (Conti et al. 2022).

At the same time, when using HVIs, there is a risk that emotional bonds and expressions will only be accepted to a limited extent by users, due to their virtual entity. The appearance and behaviour of an HVI can blur the fictional and human sides and make people feel more fearful and uncomfortable with the apparent perfection of non-human entities (Park et al. 2021). In addition, users may doubt the trustworthiness and credibility of the HVIs when advertising certain products, such as make-up or facial care, as HVIs cannot physically test these types of products (Kádeková/Holienčinová 2018, Conti et al. 2022).

A study conducted in Germany in 2023 found that 19% of all social media users were aware of HVIs. Among 16–29-year-olds, the figure was 44%. 66% of those who are aware of HVIs also follow them (respectively 76% of 16–29-year-olds). 29% of those who follow HVIs and 21% of those who know HVIs want to try out recommendations. The preferred social media channels are mainly the video-based channels Instagram and TikTok (Deckers et al. 2023).

Against this background, this paper aims to answer the following research questions:

  • What is the influence of HVIs on the purchase intention of social media users in Germany?

  • Which factors influence the intention to buy a product recommended by an HVI?

2 Conceptual Framework

2.1 Literature Review on the State of Research on Human Virtual Influencers

As a relatively recent phenomenon, HVIs have not been extensively studied from a scientific perspective. Drenten/Brooks (2020) address the contradictions between the artificiality and authenticity of HVIs and HIs, arguing that HVIs are no less real than HIs. HIs operate in an online environment of digitally processed images and filters that, like their virtual counterparts, suggest artificiality.

Robinson (2020) argues in his publication on the ontological and ethical question of HVIs that trust and thus transparency can be created by revealing the fictitious form of an HVI. The associated characteristics of the HVI are on a par with the characteristics of an HI and therefore allow products to be advertised equally successfully. Similar to HVIs, HIs are not always authentic and transparent, but create an illusion that does not exist in reality.

Moustakes et al. (2020) showed in their study that HVIs can achieve an emotional bond by elaborating a profound “life” story. Anthropomorphism in particular is important for building a parasocial relationship and can influence the success of product promotions.

Arsenyan/Mirowska (2021) and Park et al. (2021) conducted data analyses of user comments on HVIs and HIs posts. They found that there is a stronger scepticism and aversion towards an HVI. The empirical study by Cheung/Leung (2021) shows a similar result. In this study, the advertising effectiveness of HVIs was analysed in comparison to AVIs by asking test subjects to comment on products advertised by VIs. For this purpose, theoretical constructs of attractiveness, trust and expertise were used from the celebrity endorsement research. This was used to assess whether VIs are attractive, trustworthy and have the professional expertise for advertising. Significant differences were found between the VIs and HVIs in the theoretical constructs of attractiveness and trust, but not in expertise.

Batista/Chimenti (2021) identified five features that characterize an HVI. Firstly, the attractiveness of an HVI and secondly, the authenticity that is attributed to the human traits of the HVI, both in terms of physical appearance and in terms of the conceived character. Another characteristic is anthropomorphism. According to the authors, anthropomorphism can have both positive and negative effects. For example, anthropomorphism can lead to a friendly rapprochement and parasocial relationship between the HVI and its followers. However, anthropomorphism can also be the cause of a negative attitude based on his virtual entity. The final characteristics from the study are controllability and scalability. Lee/Kim (2022) prove in a study the positive effect of anthropomorphism on the parasocial relationship.

Hofeditz et al. (2022) investigated the construct of trust, which helps the HIs to advertise a company’s products. Their study shows that the majority of test subjects were unable to perceive the difference between the HI and the HVI. In order to test the theoretical constructs of trust, anthropomorphism and social presence, the identities of the HVI were revealed in the further course of the survey. The results of the study show that all three constructs are perceived significantly lower for an HVI than for an HI.

To summarise, most studies to date have focused on analysing the perception of an HVI. For this purpose, characteristics for the identification of an HVI are established, which can be derived from the appearance and behaviour of an HVI (Batista/Chimenti 2021). The characteristics have been discussed in various papers, which focus in particular on anthropomorphism with its different effects, e.g. with regard to the parasocial relationship (Moustakes et al. 2020, Batista/Chimenti 2021, Hofeditz et al. 2022, Lee/Kim 2022). The authenticity of an HVI is also controversially discussed in research (Drenten/Brooks 2020, Robinson 2020, Park et al. 2021). So far, the research findings on the perception of an HVI presented above have not addressed the influence on user behaviour, such as the intention to purchase an advertised product.

2.2 Theoretical Model and Variables

When developing the research model, in addition to the findings already mentioned in the literature review, effects from research on HIs were transferred to the HVIs research object. Studies were used that deal with the influence of HIs on the purchasing behaviour of social media users (e.g. Djafarova/Rushworrth 2017, Lou/Yuan 2018, Sokolova/Kefi 2020). In the context of this research, the question arises as to what extent HVIs can also influence the purchasing behaviour of social media users. The dependent variable of the research model is the social media user’s intention to purchase a product advertised by the HVI.

In the model developed, the purchase intention is influenced by four independent variables (see Fig. 1). These are the parasocial relationship, anthropomorphism, credibility and authenticity. Demographic characteristics such as age and gender as well as product fit were considered as moderator variables.

Several research results show that the parasocial relationship between an HI and its followers influences the purchase intention (Lee/Watkins 2016, Hwang/Zhang 2018, Sokolova/Kefi 2019, Hanief et al. 2019). Parasocial relationships arise from previous parasocial interactions (Schramm/ Hartmann 2010). In a parasocial interaction, the interaction is initiated by the media person, in this case the HI, e.g. by the HI directly addressing the audience or allowing them to participate in their lives (Cohen 1999). This creates a simulated mutual exchange (Schramm/Hartmann 2010). The parasocial relationship with the HVI is defined (as with the HI) as a social behaviour characterised by repeated parasocial interaction between HVI and social media users, which creates an interpersonal relationship.

Sokolva/Kefi (2019) showed that a parasocial interaction between the two actors has a significant positive correlation with the followers’ intention to purchase. Lee/Watkins (2016) and Hanief et al. (2019) also found that a long-term and intensive relationship between HI and follower increases the likelihood of a repeat purchase. The advertising messages of the HI are particularly effective due to the establishment of parasocial relationships (Hwang/Zhang 2018). These research findings can also be applied to the effects of the parasocial relationship between HVIs and followers on intent. Accordingly, the following hypothesis is derived: H1: The parasocial relationship has a positive influence on the purchase intention.

Fig. 1.
figure 1

Research model

Full size image

Anthropomorphism describes in the present study the human-like appearance of an HVI in facial expressions, gestures and its behaviour towards the social media user.

Research in the field of human-robot interaction has intensively investigated anthropomorphism, i.e. the attribution of human characteristics to technology, both from a technological and psychological point of view (Fussel et al. 2008, Bartneck et al. 2008). This research is based on findings from the theory “Computers are Social Actors”, according to which people react to computers with social behaviours even though they are aware that they are machines (Nass et al. 1994). People’s emotional response to robots initially becomes more positive the more human-like they look. However, this positive reaction reaches a point (“Uncanny Valley”) where small but imperfect human-like features lead to the figure being perceived as creepy or scary. Only when the character can no longer be distinguished from a real person does this feeling disappear again (Mori et al. 2012). The phenomenon of the Uncanny Valley also poses a challenge for HVIs.

Several studies suggest that brands employing anthropomorphic advertisers are often viewed more favourably by consumers (Laksmidewi et al. 2017). In addition, they are said to lead to a higher purchase intention than brands that do not use anthropomorphic advertisers (Deligoz/Ünal 2021). Further research has also shown that anthropomorphic entities create increased acceptance and familiarity with the audience and influence purchase intentions (Sheehan et al. 2020). The following hypothesis is derived from this argumentation. H2: Anthropomorphism has a positive influence on purchase intentions.

Empirical studies show that anthropomorphism is significantly related to the development of a parasocial relationships between HVIs and social media users. Both qualitative research on HVIs (Moustakas et al. 2020, Batista/Chimenti 2021) and a quantitative study by Lee/Kim (2022) showed that anthropomorphism can strengthen the development of parasocial relationships. Therefore, in addition to the direct effect of anthropomorphism on purchase intention, this study also examines the effect of anthropomorphism on parasocial relationships. H2a: Anthropomorphism has a positive influence on the parasocial relationship.

Credibility describes in this study the degree to which social media users are willing to accept the statements of the HVI as valid.

Studies in the social media context have already shown that social media users who have an intention to purchase have been influenced by the credibility of the HI (Lee/Kim 2020). Users who perceive an HI as credible are more likely to accept the HI’s product promotion, which increases the user’s purchase intention (Weismueller et al. 2020, Ata et al. 2022). As HVIs act like their human counterparts, it is assumed that HVIs can also positively influence purchase intentions through their credibility. This results in the following hypothesis. H3: Credibility has a positive influence on purchase intentions.

In this paper, credibility is conceptualised as a three-dimensional construct with the dimensions of expertise, attractiveness and trustworthiness of the advertiser (Ohanian 1990).

Expertise refers to the relevant knowledge, experience and skills that a communicator has in relation to a subject matter (Sari et al. 2021). Depending on their area of interest, HIs can be seen as experts in this field. The perceived expertise of the HI is determined by the level of experience and specialist knowledge (Weismueller et al. 2020).

Trustworthiness:

Erdogan (1999) defines trustworthiness as the honesty and integrity of an advertiser. In particular, long-term followers who view an HI as trustworthy, caring and professionally competent are more likely to buy the products presented by the HI (Sokolova/Kefi 2019).

Attractiveness:

In the context of human advertisers, attractiveness is considered to be the appeal that an advertiser exudes, which attracts the attention of consumers (Sari et al. 2021). In the HI context, the attractiveness of the influencer has an impact on social media user behaviour (Sesar et al. 2022). In particular, the physical appearance of the HI leads to new users following an HI profile on social media (Djafarova/Rushwoorth 2017).

Authenticity is seen as one of the most important factors for the success of HIs (Moore et al. 2018). Lee/Johnson (2021) describe authenticity as a characteristic held by a person who is perceived as sincere and genuine. In terms of HI advertising, authenticity has a positive effect on followers’ intention to buy (Pöyry et al. 2019). Authenticity is understood here as the extent to which a social media user perceives the content shown (e.g. the advertising of products) of an HVI as truthful and sincere. Analogue to the findings regarding HI, the following is assumed: H4: Authenticity has a positive influence on the purchase intention.

Moderator Variables.

Demographic characteristics such as age and gender as well as product fit were considered as moderator variables.

Age.

Daily use of social media applications takes place in all age groups, but mainly in the 14 to 29 age group (Beisch/Koch 2021). Studies show that younger users are more influenced by an HI (Skolova/Kefi 2019). There are also differences between 16–29-year-olds and the rest of social media users in Germany in terms of awareness of HVIs and willingness to follow them: while only 19% of all social media users are aware of HVIs, the figure for 16–29-year-olds was 44%. 76% of 16–29-year-olds who know HVIs usually follow them (66% of all social media users) (Deckers et al. 2023). Based on this behaviour, the following moderation hypothesis is formulated: H-MOD1: The positive influence of the antecedents on the purchase intention is moderated by the age of the user, such that the effect is stronger for younger users.

Gender: Studies show that women often imitate their favorite celebrities in terms of fashion and lifestyle.

This relationship is also evident among HIs and their followers (Djafarova/Rushworth 2017). An analysis of various VI Instagram profiles from 2019 showed that the core target group of HVIs consists of women (Hypeauditor 2019). Based on this behaviour, the following moderation hypothesis is formulated: H-MOD2: The positive influence of the antecedents on the purchase intention is moderated by the gender of the user, in the sense that the effect is stronger for female users.

Product Fit: Influencer marketing research examined the fit between the HI and the advertised product.

This fit describes the perceived consistency and associative connection between a brand and the advertising (Till/Busler 1998). It has been confirmed that a fit between HI and product can influence the purchase intention of followers (Belanche et al. 2021, Ju/Lou 2022). The following moderation hypothesis is formulated: H-MOD3: The positive influence of the antecedents on purchase intention is moderated by product fit, such that the effect is stronger when product fit is present.

Operationalisation.

In this paper, existing operationalisations from the literature were adapted to measure the latent constructs and adapted to the object of research. The operationalisation can be seen in Appendix 1. The items were measured using five-point Likert scales (1 = “strongly disagree” to 5 = “strongly agree”). A qualitative (3 interviews) and quantitative pre-test (33 test subjects) took place before the final data collection.

3 Research Design

The data was collected by means of a quantitative survey (standardised online survey). Instagram users in Germany aged between 15 and 60 who use Instagram at least once a week were used as the population for this research project. Instagram was chosen as the social media platform because it is one of the most important social media channels for the use of HVIs (Casaló et al. 2020). The limitation of age is based on the findings on the usage structure by age group on Instagram in Germany (ARD/ZDF-Homepage 2023). Quotas were also set in order to achieve a sufficient number of cases within each gender and age group.

The research object of this paper is the “female” HVI Lil Miquela, a “19-year-old robot living in LA”, as stated in “her” bio. Lil Miquela has been active on Instagram since 2016 and had 2.6 million followers on the channel worldwide in early 2024 (@lilmiquella). Lil Miquela behaves like a HI, posts about her leisure activities (e.g. holidays in Barcelona or visits to the hairdresser) and promotes various products and brands. Lil Miquela was created by the design studio Brud from LA (USA). Lil Miquela was chosen because it was not only “the first hyperrealistic virtual influencer to achieve mainstream popularity” and one of the most followed virtual influencers of 2022 (Virtual Humans 2024), but also because it already had advertising partnerships with several world-renowned advertising companies and brands. A selection of these brands includes Samsung, UGG, Calvin Klein, the luxury labels Chanel and Prada, and the car brands Mini and BMW (@lilmiquella).

The survey began with a detailed introduction of the HVI Lil Miquela and its profile on Instagram: Various scenarios in the areas of behaviour, leisure activities and product promotion were shown in the form of photos and videos. The HVI’s Instagram profile description and Instagram profile overview were also presented. The questionnaire consisted of 39 questions.

4 Results

4.1 Measurement Validation

The sample comprised 252 participants and was made up of 48% men and 52% women. In accordance with the quota system, two thirds of the test subjects (78%) belonged to the younger age groups (15–39 years). Only 15.5% of the test respondents had experience with HVIs in the past.

Before constructing indices for the subsequent regression analysis, it was first necessary to check the validity of the operationalised latent constructs. The measurement validation took place firstly for the first-order constructs and then for the second-order construct (credibility). This was done in a three-stage procedure according to the approach of Zinnbauer and Eberl (2004). The results (including threshold values) are shown in Table 1.

In the first step, the reliability of these constructs was tested using Cronbach’s alpha and item-to-total correlation (ITC). Reliability was confirmed for all constructs. Subsequently, the remaining items were subjected to an unifactorial exploratory factor analysis (with IBM SPSS) to determine whether the indicators represent a single construct (factor). This step also included an assessment of the reliability of the indicators (commonality) and the factor loading. The final step involves carrying out a confirmatory factor analysis with IBM SPSS AMOS. The significance of the factor loadings at the unifactorial level as well as the factor reliability (FR) and the average variance extracted (AVE) are checked first. All constructs met the criteria required in the literature with regard to reliability and convergence validity.

Following the measurement validation of the first-order constructs, the second-order quality test was carried out for the latent construct credibility and its dimensions of expertise, trust and attractiveness. The quality tests show that the constructs fulfil the minimum requirements for reliability (Cronbach’s alpha: 0.811, ITC > 0.5, commonality > 0.5, factor loading > 0.70) and validity (AVE: 0.843 and FR: 0.907). The significance of the factor loadings of the individual dimensions can also be confirmed.

Table 1. Confirmatory factor analysis results including quality criteria
Full size table

Subsequently, on a model level, discriminant validity of the 4 antecedents is examined using the Fornell-Larcker criterion: The average recorded variance of a factor should be greater than all its squared correlations with other factors (Zinnbauer/Eberl 2004). As can be seen from Table 2, this criterion is also fulfilled.

The final quality check of the overall model (with AMOS) shows that the required quality measures of the model fit are fulfilled: RMSEA = 0.084 (≥0.08), CMIN/DF = 1.943 (≤2.5) and CFI = 0.939 (≥0.9).

The indexing of the items via the average of the constructs is therefore permitted. The averages (A) of the variables are in the lower range between 1.88 and 2.63. The purchase intention of the respondents is rather low (A = 2.04; SD = 1.07). This indicates that the participants have no intention to purchase the product when it is advertised by the HVI.

4.2 Hypotheses Tests

Direct Effects.

The hypothesised relationships were examined using multiple regression analysis. The results of the regression analysis are shown in Table 3. The research model makes a highly significant contribution to explaining the purchase intention (adjR2 = 0.827; F = 162.873).

The parasocial relationship (A = 1.88, SD = 0.95) has a strong (ß = 0.468) positive and highly significant (p < 0.001) effect on purchase intention. Accordingly, the hypothesis H1 can be supported.

In contrast, anthropomorphism (A = 2.63, SD = 0.97) shows a very weak effect, which is negative contrary to the assumption (ß = - 0.063). However, this result is not significant (p > 0.05). The hypothesis H2 must therefore be rejected.

Credibility (A = 2.33, SD = 1) has a significant medium positive effect (ß = 0.223: p < 0.05) on the purchase intention, the hypothesis can be supported.

The influence of the authenticity (A = 2.35, SD = 1.10) shows a highly significant (p < 0.001) medium positive effect (ß = 0.362) on the purchase intention. Hypothesis H4 can be supported.

In addition, the simple regression analysis (adjR2 = 0.314; F = 62.79) shows a significant influence of the predictor anthropomorphism on the parasocial relationship (ß = 0.564; p < 0.001). The hypothesis H2a can be supported.

Moderating Effects.

The variables age, gender and product fit were included in the research model as moderators. To investigate the moderating effects, the overall sample was split into groups based on the hypothesised moderators (Urban/Mayerl 2011). The moderator age was split into the two groups younger (15–39 years) and older (40–59 years), the moderator gender was divided into the groups female and male and the moderator product fit was divided into the groups “non-existing” and “existing” product fit. A multi-group analysis was then carried out at sub-sample level and the difference between the groups was tested for significance using the t-test for independent samples (Table 4):

  • No significant group differences can be proven for age and the corresponding moderator hypothesis H-MOD1 must be rejected.

  • The hypothesis on the moderating effect of gender can only be supported for the effect of the parasocial relationship on the purchase intention (p = 0.033). For women, the influence of the parasocial relationship on purchase intention is stronger (B = 0.601; p < 0.001) than for men (B = 0.440; p < 0.001). With regard to the remaining predictors, the H-MOD2 hypothesis must be rejected due to the lack of significance in the group differences.

  • The analysis of the influence of the moderating effect of product fit shows significant differences in the groups for the parasocial relationship and authenticity (p = 0.000). Thus, the influence of the parasocial relationship on the purchase intention is significantly more pronounced with an existing product fit (B = 0.507; p < 0.001) than with a non-existing product fit (B = 0.429; p = 0.003). The influence of authenticity on purchase intention is also moderated by product fit. If product fit is given, there is a greater effect (B = 0.414; p < 0.001) than if product fit is not given (B = 0.264; p = 0.011). With regard to the remaining antecedents, the H-MOD3 hypothesis must be rejected due to the lack of significance in the group differences.

The tested research model can be seen in Fig. 2.

Table 2. Results of the Fornell-Larcker criterion for 2nd order constructs
Full size table
Table 3. Results of the regression analysis
Full size table
Table 4. Results of hypotheses tests for the moderating effects of Age, Gender and Product Fit
Full size table
Fig. 2.
figure 2

Results of the regression analysis

Full size image

5 Conclusion

5.1 Discussion and Business Implications

In summary, the present study confirms the model variables that were transferred from the research on HI to the HVI object of investigation. The variables parasocial relationship, authenticity and credibility have a high to highly significant influence on the purchase intention. In contrast, no direct significant influence on the purchase intention could be measured for anthropomorphism, the only indicator variable that considers the specifics of HVI´s virtuality and originates from research in the field of human-robot interaction. On the other hand, the indirect effect could be confirmed: As often mentioned in literature, anthropomorphism has a direct effect on the parasocial relationship. Another result of the present study is that the purchase intention of a product advertised by the analysed HVI is low in Germany.

Based on these findings, recommendations can be made both for the providers of HVIs and for companies wishing to use HVIs as part of their communication strategy.

For Providers of human virtual influencers (HVIs):

  • Focus on parasocial relationships: Develop strategies that foster a sense of connection between HVIs and their followers. This could involve interactive content or storytelling that resonates with the audience, enhancing the perceived closeness with the HVI. It should be noted that the various measures for building a parasocial relationship cannot be implemented in the short term, however, because a parasocial relationship only develops through repeated parasocial interactions.

  • Leverage anthropomorphism: Focus on creating HVIs with human-like attributes and the ability to establish strong parasocial relationships. To reduce negative reactions from the Uncanny Valley effect, it’s important to clearly represent HVIs as computer-generated entities, emphasizing their virtual nature.

  • Build credibility: Ensure that HVIs demonstrate expertise, trust, and attractiveness as these attributes significantly boost their credibility. This could involve showcasing their knowledge about the product or industry and creating a trustworthy persona.

  • Ensure authenticity: Maintain a high level of authenticity in the HVIs’ interactions and content. Authenticity significantly impacts purchase intentions, especially when there is a strong product fit.

  • Secure Product Fit: Align the HVI’s persona and the products they promote. A strong product fit amplifies the impact of parasocial relationships and authenticity on purchase intentions.

  • Tailor HVI strategies for different audiences: Recognize that different audience segments may respond differently to HVIs. Customize content and engagement strategies to cater to these varying preferences and responses. This also means that it is not possible to cover the entire market with a single HVI. It may well make sense to create several HVIs.

Recommendations for companies aiming to use HVIs in their communication strategies to improve engagement and influence purchase decisions:

  • Strategic HVI selection: Choose HVIs that align with your brand values and the products being promoted, ensuring a strong product fit. The HVIs should have a strong parasocial relationship with their community.

  • Focus on authenticity and credibility: Utilize HVIs that not only fit your product but also exhibit authenticity and credibility.

  • Consider gender moderation effects: Tailor strategies for HVIs keeping in mind that women might respond more strongly to parasocial relationships than men.

5.2 Limitations and Future Research

This study focussed on Germany, where there are comparatively few HVIs and therefore consumers have limited experience with them. We would expect significant further insights from testing the developed model in other countries. Further studies should also consider the influence of “experience with HVIs” as an additional factor (either as an antecedent or moderator variable). The subject of this study is the “female” HVI Lil Miquela and “her” profile on Instagram, which may limit generalisability. Future research should include different HVIs across genders, age groups and social media platforms. The influence of anthropomorphism on purchase intention should be further investigated, especially considering the Uncanny Valley effect. Also, the aspect “language and voice” will most probably play a significant role. A methodological restriction of this research is the sample size, indicating the need for larger samples in the future. Finally, an analysis comparing human influencers and virtual influencers would provide us with an indication of what the future upward potential, as well as limitations, of the virtual influencer phenomenon can be.