When “nonsense†makes sense: Effects of non-informational speech by virtual streamers in livestream commerce

Artificial intelligence has enabled the rapid adoption of virtual streamers in livestream e-commerce, making their language style a consequential determinant of selling effectiveness. Non-informational speech—utterances that add little to no new information and may seem dispensable, yet shape conversational flow—is widely viewed as a cue to naturalness and immersion in human interaction. However, its persuasive role when the source is a non-human virtual streamer, as well as the mechanisms through which it operates, has been underexplored. We distinguish two forms of non-informational speech (filler vs. redundant) and report four laboratory experiments that examine their effects on consumers' livestream purchase intention, the mediating mechanisms, and the boundary conditions. The results show that (1) The use of non-informational speech by virtual streamers will positively influence consumers' purchase intentions. Compared with redundant talk, filler talk is more effective in increasing consumers' purchase intentions. (2) social presence and attitudes toward the streamer mediate this effect; and (3) AI identity disclosure and AI familiarity serve as boundary conditions, moderating the effect of non-informational speech on purchase intention. These findings contribute to research on virtual streamers’ language styles and non-informational speech, while also offering managerial implications for firms designing marketing strategies involving virtual streamers.