Uncritical Fabulation and Holocaust Vibe: How AI Constructs the Past

Large Language Models (LLMs) produce probabilistic narratives that tend to uncritically abstract, reify, and echo the most pervasive representations of the past. Through an empirical study of AI-generated language, we show how Holocaust histories and memories created by AI appear cognitively coherent and emotionally resonant because they give rise to a new—and not unproblematic—genre that we call ‘Holocaust vibe’. Situated between the binary of authentic and artificial, Holocaust vibe is a feeling of plausibility evoked by the model’s approximate alignment with familiar representations of the Holocaust.