From Virtual Credibility to Luxury Purchase Intention: A Hybrid PLS-SEM–ANN Study of Chinese Generation Z

Virtual influencers are used to communicate luxury-brand meanings, but the process by which source attributes become purchase intention remains insufficiently established. This study integrates source credibility, meaning-transfer, and parasocial perspectives and tests paths from virtual-influencer attributes through parasocial interaction and brand attitude to purchase intention, including theory-driven serial indirect effects. Online data were obtained from 631 Chinese Generation Z respondents and after the filtering of those without virtual influencer exposure, the database retained 511 with prior virtual-influencer experience. PLS-SEM in SmartPLS 4 and artificial neural networks in R were combined. PER4, PSI3, and PSI5 were removed after first-run loadings of 0.569, 0.509, and 0.583. The final model explained 34.3% of parasocial interaction, 42.6% of brand attitude, and 44.3% of purchase intention. All specific indirect effects were positive; three serial paths were supported, while the expertise chain was equivocal because its percentile interval excluded zero but p = 0.055. Anthropomorphism strengthened the brand attitude–purchase intention relationship, while emotional engagement did not moderate the parasocial-interaction path. Repeated nested validation produced predictor rankings consistent with PLS-SEM, but ANN errors did not differ significantly from matched linear benchmarks. Relational and attitudinal conversion, rather than widespread direct persuasion or strong nonlinearity, was central to virtual luxury influence.