Post-Truth After Generation: Mechanisms Through Which Generative AI Shapes Online Public Opinion

Abstract

This study examines the role of generative artificial intelligence in online public opinion arenas. Using ethnographic observation, semi-structured interviews, and controlled experiments, it systematically analyzes how AI reshapes the workflow of public opinion production, the logic of expression, and public cognition in a post-truth context. A seven-day participatory observation shows that generative AI significantly increases production efficiency, lowers creative barriers, and reinforces content homogenization through default textual and visual templates. Interviews with 15 AI content creators indicate a motivational shift from serious viewpoint articulation toward entertainment-oriented and accommodative remixing, suggesting that AI tools not only change modes of participation but also reconstruct expressive logic. Experimental results further confirm a “viewpoint density effect”: when participants are exposed to different quantities of AI-generated content on the same topic, their attitudes, emotional intensity, information judgment, and post-truth perception all increase linearly and significantly. Overall, generative AI influences public emotion and cognitive bias at the micro level through scalable generation, stylistic discipline, and algorithm-driven feedback loops, and reshapes the public opinion ecosystem at the macro level. These findings provide key theoretical and empirical support for understanding public opinion evolution in the AI era.