Public Opinion

Millions of people around the world query large language models (LLMs) for information. Although several studies have compellingly documented the persuasive potential of these models, there is limited evidence of who or what influences the models themselves, leading to a flurry of concerns about which companies and governments build and regulate the models. Here we show through six studies that government control of the media across the world already influences the output of LLMs via their training data. We use a cross-national audit to show that LLMs exhibit a stronger pro-government valence when prompted in the languages of countries with lower media freedom than in those with higher media freedom. This result is correlational, so to triangulate the specific mechanism of how state media control can influence LLMs, we develop a multi-part case study on China’s media. We demonstrate that media scripted and curated by the Chinese state appears in LLM training datasets. To evaluate the plausible effect of this inclusion, we use an open-weight model to show that additional pretraining on Chinese state-coordinated media generates more positive answers to prompts about Chinese political institutions and leaders. We link this phenomenon to commercial models through two audit studies demonstrating that prompting models in Chinese generates more positive responses about China’s institutions and leaders than do the same queries in English. The combination of influence and persuasive potential across languages suggests the troubling conclusion that states and powerful institutions have increased strategic incentives to leverage media control in the hopes of shaping LLM output.

Large language models (LLMs) are increasingly used as synthetic survey respondents, yet it is unclear whether their belief-system structure matches that of real publics. We compare 28 LLMs to the 2024 General Social Survey (GSS) using 52 attitude items and demographic persona traits. We estimate polychoric correlation matrices and propagate un-certainty in the GSS via bootstrap resampling with multiple imputation. Constraint is measured by the variance share explained by the first principal component and by effective dependence, a determinant-based measure of global linear dependence. Across models, LLM personas exhibit substantially higher constraint than humans; conditioning on persona traits reduces constraint far more for LLMs, indicating greater demographic mediation. Projection onto a shared GSS basis further shows overemphasis of the leading dimension and missing secondary structure. These results caution against treating LLM personas as a reliable foundation for synthetic survey data generation.