AI-generated recommendations: Roles of language style, perceived AI human-likeness, and recommendation agent

Abstract

This research investigated the impact of linguistic styles on the acceptance of artificial intelligence (AI)-generated recommendations, using three experiments. Specifically, we considered the use of figurative versus literal language in ChatGPT. The findings of Study 1 indicated that figurative language positively affects visit intention, with imagery vividness serving as a mediator in the underlying process. Study 2 revealed that the effect of figurative language on ChatGPT recommendations was stronger for those who perceived the AI as human-like. Conversely, Study 3 showed that while the figurative language used by ChatGPT significantly boosted visit intention compared with literal language, this enhancement was less pronounced when recommendations were made by a human agent.