Researchers say AI chatbots are giving wrong voting information and repeating claims that look like Russian disinformation, AFP reported, in the first federal election where millions of voters turn to AI for political information.
AFP reported on November 1 that researchers have found chatbots giving wrong voting guidance, confusing candidates and echoing Russian propaganda ahead of the November 3 US midterms. Chatbots can also debunk conspiracy theories and educate voters, but the common criticism is that they often draw on unreliable web sources.
광고 문의 · 300×250AFP cited a study by Forum AI, an AI model evaluation start-up. Of eight models examined, almost 40 percent of answers to election questions contained at least one factual error, and voting-logistics questions went wrong more often than candidate questions. A representative case was Google Gemini giving November 3 as the Pennsylvania mail-ballot request deadline when the real date was a week earlier, which the study warned would leave a voter unable to use a ballot.
Russia-linked claims and the chatbots
In late October AFP fact-checkers judged a clip appearing to show a former student of Texas Senate candidate James Talarico alleging misconduct to be likely AI-altered, and AFP said the claim bore hallmarks of a Russian disinformation campaign. When NewsGuard prompted them, Google's AI Overview repeated the allegation, and Google's Gemini and Microsoft's Copilot also repeated the false claim.
Google said its systems quickly surface the latest reporting and that the issue was fixed. Microsoft said the situation was moving fast with little reliable information at the time and that Copilot later identified the claim as fabricated. A later AFP test, after the claim had been widely debunked, found the tools citing professional fact-checkers, showing answers can be inconsistent. Separately, NewsGuard studied 10 AI tools and concluded users risk being misled by foreign influence operations and other bad actors.
Voter trust and mixed verdicts
AFP said the London-based Institute for Strategic Dialogue describes this as the first US election with AI tools built into every major tech platform. A survey by the nonpartisan National Conference on Citizenship found two in five Americans had used AI to look up election information, but only 40 percent trusted it for logistics such as when, where and how to vote, and 60 percent trusted it little or not at all to be unbiased about candidates.
The picture is not one-sided. A Brennan Center for Justice study found leading chatbots rejected common election conspiracy theories, including claims that voting machines were rigged, even when users pushed back repeatedly, though the center noted they also made mistakes and tended to act as sympathetic listeners to skeptics. Anthropic, named in the Forum AI study as an example of an error, did not respond to AFP's request for comment, and Google said it focuses on helping voters find reliable information and has quality and policy protections for election topics.
