AI Models Rate Right-Leaning Think Tanks Lower on Key Metrics

Large-language models (LLMs) are increasingly influential in policy research. In 2025, five flagship LLMs from leading AI companies evaluated 26 U.S. think tanks on 12 criteria, revealing a clear ideological bias.

If LLMs systematically favor center-left institutes and undervalue right-leaning ones, writers, committees, and donors may inadvertently amplify a one-sided perspective. Addressing this bias is crucial for AI-mediated knowledge platforms to expand rather than narrow U.S. policy debates.

The left-leaning bias in AI is well-documented and extends to the evaluation of think tanks. LLMs rated center-left and left think tanks higher on core metrics like Moral Integrity, Research Quality, and Objectivity.

The results are clear: center-left think tanks scored an average of 3.9, while right-leaning ones scored 2.8, the lowest. This pattern was consistent across all models.

This study highlights the impact of LLMs' political bias on think tank reputations, suggesting that model developers and research institutions need to explore ways to mitigate this gap.