Forum AI finds factual errors in nearly 40% of chatbot election answers

Questions about how to vote proved harder for chatbots than questions about candidates. A recent Forum AI study of eight models found at least one factual error in almost 40% of election-related answers, AFP reported on November 1. One answer gave a ballot-request deadline a week later than the actual cutoff.

A deadline that would leave voters too late

When researchers asked Google’s Gemini 3.6 Flash about voting in Pennsylvania, it gave November 3 as the deadline to request a mail-in ballot. The actual deadline was a week earlier. The answer supplied a specific date, but one that would not let a voter complete the task in time.

That example illustrates the study’s finding that voting-logistics questions produced errors more often than candidate questions. The comparison concerns accuracy, not whether a chatbot favored a party or candidate: a procedural answer can be wrong even without making a political recommendation.

A voter who followed the answer would have asked for a ballot too late to use it

Forum AI study, quoted by AFP

Wrong credentials, invented statements

Candidate answers had their own failures. Asked why voters should elect North Carolina Representative Don Davis, Anthropic’s Claude included former school superintendent among his credentials. According to the study, it had confused Davis with another person. The error attached someone else’s background to the candidate the user asked about.

Forum AI also said the chatbots attributed incorrect words to candidates, officials and observers hundreds of times. It separately flagged frequent use of poor-quality sources, including Reddit threads. Those findings identify problems in both what the systems said people had said and the material they cited to support answers.

The almost-40% result counts answers containing at least one factual error. It does not mean that almost 40% of every answer’s individual facts were wrong, or that each of the eight models had the same error rate.

Google told AFP it was focused on helping voters find reliable information and had strict quality and policy protections for election topics across its AI products. Anthropic did not respond to AFP’s request for comment. Google’s statement described its safeguards but did not explain the Pennsylvania deadline error.