Human-like bots are not humans: The weakness of sensory language for virtual streamers in livestream commerce

In the last few years, virtual streamers (i.e., digital characters with human-like appearances) have been heavily utilized in the field of livestream commerce. This work examines how virtual streamers’ use of sensory language (e.g., words like “tasty” and “smooth” that engage the senses) shapes consumer responses to the sponsored products. A multi-method investigation, combining three online scenario-based experiments and one focus group, demonstrates that sensory language leads to a decrease in purchase intention. This negative effect is driven by the violation of language expectancy, which states that sensory language should not be used because bots cannot actually use and experience products. However, the effect changes from negative to positive when viewers realize that a virtual streamer is controlled by a human operator instead of an artificial intelligence program. These findings shed light on how language shapes consumer responses to virtual streamers as well as how to enhance virtual streamers’ success in livestream commerce.