AI chatbots gave outdated or wrong voting answers in recent tests
AI chatbots gave voters outdated or incorrect information about voting in tests reported by NPR on November 2, 2026. The tests, run by Innovating for the Public Good, asked seven major chatbots real voter questions. The chatbots were Google AI Mode, AI Overviews, Anthropic's Claude, Microsoft Copilot, Google Gemini, OpenAI's ChatGPT and Perplexity.
In October, 13.9% of the chatbot answers to questions about how and when to vote were outdated or inaccurate. In 5.9% of answers, the errors could lead a voter to miss a deadline, vote on the wrong day or skip a required step. Across all the questions tested, 5.1% of answers contained at least one clear factual error, and 10.9% contained a clear error or a misleading claim.
Claude and Copilot produced three-quarters of the clear errors the researchers found. On voting questions, Claude gave wrong or outdated information one time in three, Copilot one time in four and ChatGPT one time in 10. Claude gave voter registration deadlines for Michigan's 4th congressional district, one of the state's U.S. House districts, that were later than the actual online and mail deadlines.
Microsoft said it has made improvements to how Copilot responds to election-related questions and is continuing to test and refine them. Some chatbots also provided lists of candidates to vote for, according to the same reporting.
Other news organizations have documented similar problems. The Washington Post tested six AI chatbots by asking them to write voter guides. When the guides specified a district, the district was wrong 22 percent of the time. A review by Futurism of 3,000 AI-generated voter guides found that top chatbots gave wrong basic voting information at rates as high as 48 percent.
Reporting by Barron's said AI chatbots had misidentified candidates in responses about U.S. elections and had repeated Russian propaganda. The Japan Times reported that chatbots had put incorrect words in the mouths of candidates, officials and observers hundreds of times. In an earlier test reported in August, popular chatbots mostly debunked falsehoods spread by other countries, according to NPR.
The use of chatbots is widespread. A Pew Research Center survey from earlier this year found about half of American adults reported using AI-powered chatbots. Of those users, 42% said they use chatbots to search for information. Reuters has reported that U.S. political candidates are tailoring their messages for the AI models voters use.
