US politicians tailor their messages to AI models used by voters

Political candidates in the United States are engineering their campaign messages to guide AI chatbots that voters frequently use to get information about issues. Strategy documents and presentations reviewed by Reuters reveal efforts by candidates to control AI outputs.

According to strategy documents that emerged in Washington, a PowerPoint presentation prepared for Mike Minogue, running as a Republican opponent against Massachusetts Governor Maura Healey, includes a strategy aimed at having chatbots present him as the preferred choice on key issues. As AI increasingly becomes an important news source, political campaigns are stepping up their efforts in this area, paying anywhere from hundreds to tens of thousands of dollars monthly for AI strategy and monitoring work, with some performance bonuses reaching $200,000.

The presentation outlines the generative engine optimization (GEO) method, where campaigns purchase space on high-authority websites cited by AI models and generate content on YouTube and Reddit. According to Cameron Hickey, about 25 percent of voters asked an AI chatbot a question about a political candidate when making their voting decision. Other firms like CampSight also offer tools that help candidates track and influence what large language models say about them.