A double-edged sword: Exploring the effectiveness of expressed subjectivity in virtual influencer endorsements

While marketers increasingly deploy linguistic anthropomorphism in avatar marketing, the effectiveness of virtual influencers' (VIs) subjectivity and the underlying mechanisms are not well understood in brand endorsements. Based on the theory of ambivalence and persuasion literature, we explore whether, why, and when VIs' expressed subjectivity influences endorsement effectiveness. Four main experiments and three supplementary experiments spanning hospitality contexts were conducted, using t-test, ANOVA, and PROCESS macro analyses. The results indicated that VIs' expressed subjectivity was less effective than that of human influencers in terms of purchase intention and brand interest, and this can be explained by a dual mediation pathway: empathy and skepticism (Studies 1–2). Furthermore, expressed subjectivity became more effective when a virtual influencer was mimic-human or controlled by a human (Studies 3–4). Our findings extend previous literature on virtual influencer marketing by unraveling the persuasion mechanism underlying the effectiveness of VIs’ expressed subjectivity. Also, we contribute to the machine anthropomorphism literature by revealing the interrelationships among linguistic and visual anthropomorphism. Additionally, this paper provides practical implications for marketers and creators who are interested in deploying VIs.