MIT researchers have launched the LLM Election Observatory dashboard to capture how leading AI chatbots answer questions about the 2026 U.S. midterm elections in real time. The dashboard simulates nationwide user‑bot interactions using a fixed questionnaire that spans election certification, hot‑button issues, result forecasts, news coverage, and more. Every few days the same set of prompts is re‑run, allowing the team to track how responses evolve with the campaign timeline, user location, party affiliation, and self‑described identity.\ \ The primary aim is to create a systematic, publicly available record of this information environment, enabling analysis of model‑to‑model differences, temporal shifts, source citations, and framing effects. Early observations show that even when factual content is identical, models vary in presentation, cited sources, and tone, potentially shaping voters’ perceptions of the same facts.\ \ Researchers advise voters that a chatbot’s confident tone does not guarantee accuracy; answers should be treated as a starting point and cross‑checked against reliable sources. At the same time, evidence‑based dialogues have shown promise in debunking misinformation at scale, and independent audits like this help assess AI’s real impact on electoral discourse.\ \ Review