MIT Launches 2026 Election Dashboard to Track AI Chatbot Answers in Real Time
Updated
Updated · MIT News · Oct 30
MIT Launches 2026 Election Dashboard to Track AI Chatbot Answers in Real Time
3 articles · Updated · MIT News · Oct 30
Summary
MIT researchers have opened the LLM Election Observatory, a public dashboard that repeatedly tests major chatbots on 2026 U.S. midterm questions and shows how their answers change over time.
Every few days, the team reruns a fixed set of prompts on topics such as election certification, result predictions and hot-button issues, varying user location, party identity and other traits.
Early 2026 data already show differences in tone, framing and cited sources across user identities even when underlying facts are similar, though the researchers said they have not completed a full longitudinal analysis.
The project builds on 2024 election research that found meaningful variation across models and prompt framing, and aims to create a public historical record of otherwise private chatbot political information.
MIT researchers said the broader goal is to assess how AI shapes voter understanding and discourse—not just factual accuracy—as chatbots become a common starting point for election information.