“Human-like or cartoon-like?” The impact of virtual influencer style on consumer forgiveness propensity following service failures

Purpose Virtual influencers have increasingly become a critical force in shaping consumer behavior. Accordingly, an in-depth examination of consumers’ response mechanisms following service failures by virtual influencers has emerged as an important research topic within the virtual influencer literature. Design/methodology/approach Using virtual influencer service failures as the research scenario, three experiments were conducted to examine consumers’ forgiveness propensity following service failures involving different types of virtual influencers. Findings The results of the three experiments show that, in service failure situations, compared with anthropomorphic virtual influencers, cartoon-like virtual influencers induce a higher level of consumer forgiveness propensity. This effect is mediated by responsibility attribution. Anthropomorphic virtual influencers are perceived as bearing greater responsibility because of their more salient human-like characteristics, which in turn, reduce consumers’ forgiveness propensity. The presence or absence of an algorithmic explanation affects consumers’ responsibility attribution for service failures, and uncanny perception negatively affects forgiveness propensity. Research limitations/implications This study focuses on recommendation failures and value-based failures and uses samples drawn from a specific cultural context. Future research could examine additional types of service failure, explore cross-cultural differences, and investigate dynamic service recovery processes over time. Practical implications The findings provide guidance for firms in selecting appropriate virtual influencer designs, managing algorithmic transparency, and developing effective service recovery strategies. Organizations should carefully balance anthropomorphic design and perceived uncanniness while tailoring communication strategies following service failures. Originality/value This study expands theoretical research on human–computer interaction and human-computer collaboration in the field of AI service marketing, while also providing practical insights for enterprises seeking to effectively optimize the remedial measures for virtual influencer service failures.