Artificially influential: examining the impact of virtual influencers on non-followers’ behavior

Purpose This study aims to address the limited understanding of how unfamiliar non-followers respond to virtual versus human influencers encountered through algorithmically curated feeds. This study tests whether disclosure that a seemingly human influencer is AI-generated lowers perceived authenticity, producing negative expectation disconfirmation that reduces engagement and purchase intentions. Design/methodology/approach Two between-subjects online experiments with US non-followers (n = 303 and n = 316) were conducted through CloudResearch Connect. Measurement validity was assessed using confirmatory factor analysis in SmartPLS 4, and hypotheses were tested using PROCESS serial mediation analyses with bootstrapping. Findings Across both studies, virtual influencer disclosure significantly lowered perceived authenticity, which generated more negative expectation disconfirmation and reduced engagement intentions. In Study 2, this serial process extended to lower purchase intentions. The hypothesized serial indirect effects were significant. Practical implications Brands using human-presenting virtual influencers with unfamiliar audiences should treat disclosure as expectation management by clearly identifying the influencer as AI-generated or virtual from initial exposure. Managers should also recognize that disclosure may lower perceived authenticity, engagement and purchase intentions. Originality/value This study examines an underexplored stage of virtual influencer exposure involving unfamiliar non-followers. This study broadens the expectation–performance comparison process of expectation disconfirmation theory by showing that, during initial encounters, the comparison standard can derive from category-based expectations formed without prior experience with the focal influencer. Perceived authenticity serves as the evaluative mechanism linking newly revealed information to disconfirmation.