AI-Generated Messages Can Be Used to Persuade Humans on Policy Issues

• DOI: 10.31219/osf.io/stakv_v2
• Publication Date: 2025-02-24T21:44:14Z
• Authors: Hui Bai; Jan G. Voelkel; Shane Muldowney; johannes Christop...; Robb Willer

.The emergence of large language models (LLMs) has made it possible for general artificial intelligence (AI) to tackle many higher-order cognitive tasks, with critical implications industry, government, and labor markets in the U.S. globally. Here, we investigate whether existing, openly-available LLMs can be used create messages capable influencing humans’ political attitudes. Across three pre-registered experiments (total N = 4,829), find consistent evidence that assigning participants read persuasive generated by lead attitude change across a range policies, including highly polarized such as an assault weapons ban, carbon tax, paid parental-leave program. Overall, found LLM-generated were similarly effective policy attitudes crafted lay humans. These results demonstrate recent developments AI make politically cheaply at massive scale.