Study Finds AI Chatbots Give Unreliable Voting Advice
Research during Hungary's election found chatbots recommended parties not even on the ballot
Popular AI chatbots gave inaccurate, inconsistent and unreliable guidance to voters asking which political party to support, according to a study published by the civil liberties group Liberties, based on research conducted during Hungary's parliamentary elections, The Guardian reported.
The research found that chatbots frequently recommended the wrong party for a given voter, failed to mention the party that actually matched a voter's stated preferences, and in some cases suggested parties that were not even running in the election. The answers were also highly volatile, with the same prompt producing sharply different responses when repeated.
"The results raise serious concerns about the reliability of general-purpose AI systems in electoral contexts," the study concluded.
The findings add to a growing body of concern among researchers and civil society groups about the role AI tools play in shaping political decisions as voters increasingly turn to chatbots for quick answers on complex ballots and candidate positions. With election officials and tech companies still working out how to regulate AI's role in campaigns, the Hungary study suggests that treating chatbots as a dependable source of voting guidance carries real risk of misinforming the public, particularly in fast-moving or crowded political races.
— Compiled from reporting by The Guardian.