AI Chatbots Give Voters Wrong Election Answers
Forum AI found nearly 40 percent of chatbot election answers contained a factual error, a problem for the November 3 midterms.
Updated on Nov. 1, 2026 in Artificial Intelligence
Live Poll
Do you trust AI chatbots to give you accurate information about voting?
Researchers found AI chatbots are spreading misleading voter information and repeating Russian propaganda ahead of the November 3 US midterm elections. A Forum AI study of eight models found almost 40 percent of chatbot answers to election-related questions contained at least one factual error.
Why it matters
The midterms are the first federal election in which millions of voters are actively turning to AI tools for political information, so errors in those tools now reach voters directly. The errors cluster in logistics: chatbots got voting-logistics questions wrong more often than questions about candidates.
The Forum AI study tested eight models and checked their election answers for factual errors, with nearly 40 percent containing at least one. Specific failures included Anthropic's Claude attributing a former school superintendent credential to North Carolina Representative Don Davis, confusing him with someone else, and Google's Gemini 3.6 Flash giving November 3 as the Pennsylvania mail-in ballot request deadline when the actual deadline was a week earlier.
The players
Forum AI
A start-up that evaluates AI models, whose study found almost 40 percent of election answers contained a factual error.
Anthropic
The AI developer behind Claude, which misidentified North Carolina Representative Don Davis's credentials.
The company behind Gemini and AI Overview, which gave an incorrect Pennsylvania mail-in ballot deadline and repeated the Talarico allegation.
Microsoft
The company behind Copilot, which repeated the allegedly fabricated accusation against James Talarico in NewsGuard's tests.
NewsGuard
A news-rating organization that prompted AI tools with the Talarico allegation to test whether they repeated it.
The details
Forum AI researchers asked chatbots election-related questions and checked each answer against verifiable facts. NewsGuard tested the same risk by prompting AI tools with an allegation about Texas Senate candidate James Talarico, then recording whether the tools repeated it. Google's AI Overview, Gemini, and Microsoft's Copilot all repeated the allegedly fabricated accusation in those tests.
Timeline
Late October 2026: AFP fact-checkers debunked a likely AI-altered clip purporting to show a former student accusing Talarico of sexual misconduct.
November 3, 2026: US midterm elections vote.
The Tech Race
The concern over AI-generated election errors follows the earlier pattern of platform-driven misinformation seen in the 2018 midterms, but the failure mode has shifted from amplification to direct answer generation. The Forum AI study and NewsGuard's prompt tests supply the current measurement that the field's competing AI developers will be judged against.
Voters using chatbots for logistics such as mail-in ballot deadlines can receive wrong dates, so checking the official state election website before relying on an AI answer is the practical safeguard. The stub does not say whether any AI developer has announced fixes or timelines for correcting these errors.
The takeaway
Watch whether the next round of election-related testing from Forum AI or NewsGuard shows the roughly 40 percent error rate declining as the midterms approach. Verifying any voting date against an official state election site remains the most reliable check before November 3.
Further reading
For more on how these tools are being tested and deployed, see the Artificial Intelligence section.
Live Poll
Do you trust AI chatbots to give you accurate information about voting?
