Article Summary

An Israeli company tested how AI chatbots respond to political questions by creating detailed user profiles and asking leading chatbots who people should vote for. The study found that chatbots avoid direct political recommendations but can be steered toward them using indirect questions, and most of their responses rely heavily on news articles, raising concerns about bias if the sources are flawed. The article warns users not to trust AI for voting advice, highlighting that these systems repeat existing content rather than offering independent, intelligent guidance.