Generative artificial intelligence (AI) possesses remarkable capabilities in producing diverse forms of human-like artistic creative content. However, based on the theoretical perspective of algorithm aversion in existing literature, it appears that people do not accord AI-generated works the same level of recognition as they do to human-created works. More importantly, the psychological mechanisms underlying this bias remain vague. Based on two pilot studies and five formal experiments (four were pre-registered), the present research systematically examined the “AI-label effect” and its psychological underpinnings. Our findings revealed a persistent evaluative bias against AI-generated artworks, moderated by four key mechanisms: individuals with more favorable attitudes towards AI exhibited mitigated bias (Study 1), perceived effort positively correlated with evaluative favorability (Study 2), existential threat perceptions induced by AI exacerbated bias (Study 3), and people paid less attention to the emotional aspects of AI paintings, which led to greater bias (Study 4). Notably, this bias diminishes when evaluative criteria shift from subjective artistic considerations to objective scientific parameters (Study 5). These insights advance our understanding of human–AI interaction dynamics by elucidating the cognitive architecture of algorithmic bias, while providing empirically grounded guidance for (a) developing AI systems that align with human evaluative frameworks, and (b) designing intervention strategies to mitigate perceptual asymmetries in human–AI collaboration. • Positive attitudes towards AI reduce bias against its artworks. • More perceived effort increases people’s favorability towards AI-labeled artworks. • Perceived threat caused by AI amplify people’s bias against its artworks. • Lower emotional focus on AI fuels bias against its artworks.
Zhang et al. (2026) studied this question.