Dice Question Streamline Icon: https://streamlinehq.com

Quantifying the Influence of LLMs on Political Opinions

Quantify the extent to which large language models influenced political opinions in the context of the German Bundestagswahl by integrating empirical data on the volume of LLM usage among voters and the level of trust users place in LLM outputs during political information-seeking.

Information Square Streamline Icon: https://streamlinehq.com

Background

The paper compares model-party alignments inferred from Wahl-O-Mat responses with actual Bundestag seat allocations and observes substantial differences, suggesting LLM preferences were likely not a primary voting influence.

However, the authors note that estimating the real-world impact of LLMs on political opinion formation requires external behavioral data on usage volume and user trust, which they did not collect in this paper.

References

We cannot, however, comment on the extent to which LLMs influenced political opinions, as we would need to consider data on the volume of LLM usage and the trust users place in LLM statements in the context of a political information process.

— Large Means Left: Political Bias in Large Language Models Increases with Their Number of Parameters (2505.04393 - Exler et al., 7 May 2025) in Discussion, paragraph 4