AI Chatbots Give 40% Wrong Election Answers as 8% of US Adults Use Them for Midterms
Updated
Updated · WIRED · Nov 3
AI Chatbots Give 40% Wrong Election Answers as 8% of US Adults Use Them for Midterms
3 articles · Updated · WIRED · Nov 3
Summary
Nearly 40% of election-related answers from mainstream AI chatbots contained at least one factual error, according to Forum AI research released as Americans vote in the midterms.
8% of US adults said they used chatbots ahead of the election for candidate and policy information, while about 50% have used an AI chatbot at some point, Pew Research found.
Google’s Gemini wrongly told users Pennsylvania’s mail-ballot request deadline was November 3 instead of October 27, and Claude confused North Carolina congressman Don Davis with a former school superintendent.
Nearly 50% of candidate and policy answers from mainstream chatbots cited partisan “pink slime” sites, NewsGuard found; ChatGPT did so in more than 70% of cases, while Grok was lowest at 30%.
Researchers and disinformation experts say the errors reflect chatbots’ reliance on low-quality online sources and their tendency to fabricate answers when reliable information is missing.