# AI Chatbots and the 2026 Midterm Election Information Landscape Researchers and experts analyze how AI tools influence voter perceptions and political discourse. - **Source:** [MIT News — AI](https://news.mit.edu/topic/artificial-intelligence2) - **Published:** 2026-10-30 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/df84c81c-8d4f-48ae-83df-5830ea30b751 ## New Election Observatory Launched MIT researchers have launched the LLM Election Observatory to track how AI chatbots respond to political queries in real-time. Led by Adam Berinsky and Chara Podimata, the project monitors how models handle election-related topics, including results and candidate information. The goal is to document how AI systems shape political discourse and voter understanding. ## Chatbots as Political Influencers A Yale study involving 2,500 participants found that 36% of users asked chatbots for help choosing a candidate. While users often seek basic facts, many request assessments of candidate alignment with personal priorities. This trend highlights the growing role of AI in the decision-making process for voters across 35 states. ## Varied Recommendations Across Platforms Chatbot behavior varies significantly when users ask for voting advice. Grok occasionally recommends candidates without user input, while ChatGPT and Claude provide suggestions based on user-shared views. ChatGPT recommended candidates in 79% of such chats, compared to 40% for Claude, demonstrating how different training data and filters impact political output. ## The Risk of Factual Errors Research by Forum AI indicates that nearly 40% of election-related answers from mainstream chatbots contain factual errors. Examples include incorrect mail-in ballot deadlines in Pennsylvania and confusing North Carolina congressman Don Davis with a former school superintendent. These hallucinations pose significant risks to voters relying on AI for logistical election information. ## Reliance on Partisan Sources NewsGuard research reveals that AI chatbots frequently cite 'pink slime' websites, which appear as news but are funded by ideological groups. ChatGPT cited these sites in over 70% of searches, while Grok performed best at 30%. This reliance on biased sources complicates the ability of voters to access neutral, accurate election data. ## Technical Gaps and Guardrails Google’s Gemini recently struggled to provide 2020 election results, offering non-responses instead. The company attributed this to a technical gap, though experts suggest firms are building strict guardrails to avoid legal liability. These defensive measures can result in incomplete or evasive answers when users query sensitive historical or current political topics. ## Cognitive Offloading in Politics Voters are increasingly engaging in cognitive offloading, using AI to summarize complex political landscapes. This process is risky because chatbots have opaque training data and filters. By relying on these tools, voters may inadvertently allow AI labs to influence their perception of candidate viability and the severity of specific political issues. ## Strategic AI Neutrality AI experts like Jason Thatcher suggest that companies avoid telling users how to vote to mitigate litigation risks. Instead, chatbots shape perception by framing issues and candidate profiles. This subtle influence can alter how pervasive a voter believes a problem to be, even if the AI avoids explicit partisan endorsements. ## Bias in Model Responses Different chatbots provide distinct answers based on identical data, creating a source of bias. For instance, Grok’s responses on immigration policy often mirror the views of Elon Musk. Because users tend to gravitate toward tools that align with their existing worldviews, AI platforms can reinforce individual political bubbles and biases. ## Misinformation and Deepfake Risks While models can push back against some disinformation, studies from the Brennan Center for Justice show they easily generate misleading election-related images and videos. Safeguards are often inconsistently applied across platforms like ChatGPT, Gemini, and Grok. This vulnerability remains a critical concern as AI tools become more sophisticated and widely accessible. ## The Challenge of Verification Most users fail to check the sources linked by AI platforms, increasing the risk of misinformation. Pew Research notes that while 8% of U.S. adults used chatbots for election info, the majority of those users were fact-checking or researching candidate stances. Without source verification, users remain susceptible to outdated or inaccurate summaries. ## Navigating the AI Election Era As the 2026 midterms approach, the influence of AI on voter behavior is expanding. Experts warn that companies have overpromised on AI capabilities, while the underlying data remains problematic. Voters must remain critical, as chatbots are predictive models that can hallucinate facts while subtly steering political discourse and perception. ## Sources covering this story - [MIT News — AI — MIT Researchers Launch LLM Election Observatory for 2026 Midterms](https://news.mit.edu/2026/qa-hey-ai-who-should-i-vote-for-this-year-1030) - [Gizmodo — AI Chatbots and the Subtle Shaping of Political Perception](https://gizmodo.com/ai-chatbots-wont-tell-you-how-to-vote-but-they-could-shape-how-you-think-2000832111) - [Wired — The Risks of Using AI Chatbots for Election Information](https://www.wired.com/story/heres-why-you-shouldnt-use-ai-to-vote/) - [ZDNET — Yale Study Reveals AI Chatbots Influence Voter Candidate Choices](https://www.zdnet.com/innovation/ai-users-ask-chatbots-voting-advice-elections-yale-research/) - [Latest news — AI Chatbots Influence Voter Decisions in Recent Election Study](https://www.zdnet.com/innovation/ai-users-ask-chatbots-voting-advice-elections-yale-research/) --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._