LLMs Show Critical Acclaim Bias in Film Preferences, Study Finds

A recent study published on arXiv (arXiv:2608.06955) investigated eight models developed by Anthropic, OpenAI, Alibaba, and Mistral, focusing on their film preferences. Using a benchmark of 200 films and conducting 20,000 pairwise comparisons for each model, researchers applied Bradley–Terry estimation to reveal a distinct bias towards critically acclaimed yet commercially obscure films. This finding highlights a significant orientation in AI models that could lead to cultural bias in artificial intelligence systems, raising important questions about the influence of such preferences in broader media consumption.