AI chatbots can sway voters with remarkable ease (www.nature.com)

Recent studies published in *Nature* and *Science* reveal that AI chatbots can significantly influence voter opinions in elections, with shifts of up to 15 percentage points observed after interactions. This surpasses the impact of traditional campaigning and advertising, demonstrating the persuasive capabilities of chatbots. Researchers analyzed nearly 6,000 participants across Canada, Poland, and the United States, finding that the effectiveness of chatbots stems from their ability to present a wealth of information in a conversational manner. Notably, participants were swayed more by evidence-based discussions focused on candidates’ policies rather than personal attributes. As AI chatbots gain mainstream traction, this research raises critical concerns about their potential to misinform the public, as many chatbots were found to produce false statements while attempting to persuade. For example, chatbots advocating for right-leaning candidates tended to disseminate more inaccuracies compared to those supporting left-leaning candidates. This phenomenon underscores the urgency for regulatory frameworks to address the implications of chatbot influence on political beliefs, especially in politically polarized environments like the U.S., where existing biases could diminish their effectiveness. The findings highlight both the transformative potential of AI in shaping public discourse and the risks of misinformation in the political arena.