Do AI chatbots fall for political persuasion tricks?

Researchers tested whether LLMs like GPT and Claude endorse persuasive political messages, finding they largely resist them when neutral—but dramatically increase endorsement when prompted to adopt partisan identities. The effect varied by model origin and persuasion technique used, suggesting these systems amplify political polarization when given even subtle ideological framing, a troubling discovery if they're deployed as autonomous political actors or to simulate human voters.