Chatbots Mislead Voters With 13.9% Errors

Almost fourteen percent of AI chatbot answers on voting logistics are wrong, and that's just the tip of the iceberg. IFPG's October test covered seven major bots like ChatGPT, Claude, Copilot, Gemini, Google AI Mode and Perplexity. Overall, 13.9% of logistics answers were outdated or inaccurate. Five point nine percent could actually mislead someone about deadlines or polling places. Claude and Copilot alone generated three-quarters of the clear factual errors. Claude gave wrong or stale info about a Michigan 2026 race one in three times. Copilot missed the mark roughly one in four, while ChatGPT slipped once every ten queries. Tim Chambers warned the error rates are 'extremely troubling' for democracy. Meanwhile, Pew finds about half of U.S. adults now chat with AI, and 42% rely on it for political info. Experts say campaign operatives are already using generative engine optimization to push their narratives into bot answers. If voters trust those bot replies, the whole election process could be derailed.