Prompting the Past: Generative AI and Holocaust History 

Funded by the Claims Conference and German Ministry of Finance, this project examines how generative AI systems represent the Holocaust across multiple languages and modalities. We assess the reliability, completeness, and nuance of AI-generated responses, documenting patterns of hallucination, over-generalization, and what we call "uncritical fabulation," the tendency of AI to flatten complex histories into familiar, affectively smooth narratives. This work contributes to an emerging critical framework for understanding the impact of AI on historical knowledge and public education. A central deliverable of this project involves the creation of a platform to archive and analyze LLM responses to historical prompts across models and over time.