LLMs Generate Kitsch

This paper proposes that Large Language Models (LLMs) systematically generate kitsch as a consequence of their training method. Empirically, the study shows readers perceive LLM-generated stories as kitschier, with implications for future study design and creative tasks.