Q&A: Hey AI, who should I vote for this year?

An MIT research team has developed a new dashboard called the LLM Election Observatory to monitor how artificial intelligence chatbots respond to questions about the 2026 U.S. midterm elections. The project, led by Professor Adam Berinsky, Assistant Professor Chara Podimata, and Professor Charles Stewart III, aims to track how different AI systems answer a fixed set of questions on various topics, including election certification, hot-button issues, result predictions, news coverage, and more.

The researchers want to understand not only how AI systems spread information but also how they shape political discourse and voter understanding of political issues across the country. The dashboard runs the same set of questions through AI chatbots at regular intervals, monitoring how their responses vary depending on factors such as user location, political party affiliation, identity, and other considerations.

The goal of this research is to systematically document the evolving information environment surrounding AI chatbots and their impact on political discourse. While factual accuracy is essential, the researchers also emphasize the need to examine subtler ways in which chatbots may influence people's understanding of political issues.

Podimata, who previously studied the influence of LLMs during the 2024 election, notes that the 2026 midterms offer a unique opportunity to expand this research. Midterm elections include a multitude of state and local contests alongside broader national narratives. The updated LLM Election Observatory survey asks questions about specific contests, as well as broader issues related to voting information and the election information environment.