Mapping Geopolitical Bias in 11 Large Language Models: A Bilingual, Dual-Framing Analysis of U.S.-China Tensions

This study presents a quantitative instrument for mapping geopolitical bias in 11 large language models (LLMs) using a bilingual approach, analyzing 19,712 responses to assess U.S.-China tensions. The research reveals that model origin, query language, and issue domain significantly influence bias, with all models showing a Pro-China lean in Mandarin. The findings are made accessible through an open, interactive tool, providing practitioners with a framework to evaluate bias in LLMs across various contested opinion domains.