AIn’t it real? Using textual paralanguage to identify fake AI-generated electronic word-of-mouth
(A2024-118289)

AUTHORS

Tobias Maiberger, Darmstadt University of Applied Sciences

ABSTRACT

The rise of AI-generated electronic word-of-mouth (eWOM) is a new challenge for online platforms, brands, and consumers, putting valuable content such as online reviews at risk of losing influence. Using a theory-based approach, this study proposes to use instances of textual paralanguage (i.e., non-verbal parts of speech) in eWOM as an indicator to identify fake AI-generated eWOM. Analyzing a dataset of more than 40,000 texts reveals that eWOM containing textual paralanguage is more likely to be created by humans than by AI. By transferring insights from nonverbal communication to the realm of AI-generated eWOM, this research contributes to the understanding of what makes communication humanlike. The results add to the knowledge of existing algorithms of review platforms, offer new concepts for brands to communicate in a more human-like way, and help consumers better recognize AI-generated eWOM.