Your expression matters: the impact of linguistic content characteristics of algorithm output on consumers’ algorithm aversion and information adoption behavior
Y. Li, Qixuan Liu · Asia Pacific Journal of Marketing and Logistics · 2026
Purpose Language is the core medium of human-AI algorithm interaction, yet how its characteristics shape algorithmic decision-making outcomes remains understudied. Drawing on social presence theory, this study categorizes characteristics of AI-generated linguistic content into two dimensions: affective social presence and cognitive social presence. We further examine the moderating role of consumers’ positive emotions in mitigating algorithm aversion. Design/methodology/approach This study collected a total of 569 valid samples using a structured questionnaire, and used PLS-SEM to analyze the path relationships between various constructs in the first-order and second-order structural equation models. Findings The results show that: (1) Both affective social presence and cognitive social presence can significantly inhibit consumers’ algorithm aversion and promote information adoption behavior, with affective social presence exerting a stronger effect. (2) Anthropomorphic characteristics of the linguistic content output by AI algorithms outperform other linguistic characteristics in mitigating algorithm aversion and driving information adoption behavior. (3) Consumers’ positive emotions not only directly promote information adoption behavior, but also significantly attenuate the negative relationship between algorithm aversion and information adoption behavior. Originality/value This study extends the application of Social Presence Theory to algorithm aversion research, and offers actionable implications for AI algorithm developers and e-commerce platforms.