Who Needs Humans? A Study on How Virtual Influencer Autonomy Builds Synthetic Parasocial Bonds Stronger Than Anticipated

Virtual influencers are increasingly used in social media marketing, yet their ability to drive consumer behavior remains theoretically ambiguous. Drawing on parasocial relationship theory and human–AI interaction research, this study examines how perceived virtual influencer likeness and perceived agent autonomy shape purchase intention through synthetic parasocial relationships (S-PSR). We further account for ambivalent affective responses, coolness and eeriness, and test the moderating roles of product type and authenticity skepticism. Survey data from 372 U.S. Instagram users exposed to virtual and human influencer endorsements across technology products and experiential services were analyzed using PLS-SEM. Results show that S-PSR is the strongest predictor of purchase intention. Perceived agent autonomy significantly strengthens S-PSR, while perceived virtual likeness simultaneously increases coolness and eeriness, revealing an ambivalent pathway. The impact of S-PSR on purchase intention is stronger for technology products and weaker under high authenticity skepticism. Overall, the findings demonstrate that virtual influencers are most persuasive when perceived as autonomous, cool, rather than eerie, and contextually aligned with technology-oriented products.