This study investigates how key characteristics of AI-generated virtual streamers – perceived authenticity, anthropomorphism, social presence and interactivity – influence consumer patronage intentions in live-streaming commerce. It further examines the dual mediating roles of attractiveness and perceived cognitive fluency and explores the moderating effects of persuasion knowledge and algorithmic legitimacy to provide a comprehensive understanding of consumer responses to virtual streamers.
Two cross-national experiments were conducted with 935 participants from China and Japan. Participants viewed a simulated live-commerce video featuring an AI-generated virtual streamer. Structural equation modeling was used to test the hypothesized relationships, including direct effects, dual mediating pathways (attractiveness and cognitive fluency) and the moderating roles of persuasion knowledge and algorithmic legitimacy.
All four virtual streamer characteristics significantly influenced patronage intention through the dual mediation of attractiveness and perceived cognitive fluency. Persuasion knowledge weakened these effects, while algorithmic legitimacy strengthened them, highlighting the contrasting roles of cognitive resistance and ethical validation in shaping consumer responses. Interactivity showed unique robustness, maintaining its influence on cognitive fluency even under high persuasion knowledge.
This study extends media richness theory to AI-generated virtual streamers by uncovering how key richness cues influence consumer responses through emotional and cognitive pathways. It also contributes to persuasion knowledge research by identifying its moderating role in shaping these effects. The inclusion of algorithmic legitimacy as an ethical moderator adds a novel dimension, highlighting how normative evaluations of AI systems shape engagement in live-streaming commerce.
