The rapid adoption of digital human streamers in livestream commerce has reshaped how consumers interact with AI-driven marketing agents. While hyper-realistic digital humans are increasingly deployed to address labour shortages, extend livestreaming hours, and enhance promotional effectiveness, it remains unclear whether higher human-likeness strengthens or weakens consumers’ purchase intention. Drawing on anthropomorphism theory and the Uncanny Valley framework, this study examines how hyper-realistic human likeness influences purchase intention both directly and indirectly through perceived marketing effectiveness and affective responses. Using a survey experiment featuring digital humans with varying levels of realism, data were collected from 316 Chinese livestream commerce users and analysed using PLS-SEM. Results show that hyper-realistic human likeness does not directly increase purchase intention. Instead, its effects operate through two distinct pathways. First, greater realism enhances perceived marketing effectiveness—reflected in perceived assortment, price advantages, and channel transparency—which strongly predicts purchase intention. Second, higher realism significantly reduces perceived eeriness, indirectly improving purchase intention, although perceived uneasiness shows no meaningful influence. These findings advance livestream commerce research by demonstrating that hyper-realism functions as an indirect catalyst rather than a direct persuasive cue. The study provides strategic guidance for firms designing AI-driven livestreaming solutions and deploying hyper-realistic digital human streamers.
Liu et al. (Sun,) studied this question.