Populist Rhetoric Engages but Does Not Persuade: Evidence from 90 Countries and Three Experiments

Populism is blamed for many of democracy’s troubles, in part because its rhetoric is thought to win over mass audiences, especially online. We test that premise and find little support for it. In more than 28 million Facebook posts from over 26,000 candidates and 1,600 parties across 90 countries, posts attacking elites attract far more likes, comments, and shares than other posts—but the reactions they provoke turn negative, with anger rising and affection falling. Adding appeals to ordinary people, the ingredient said to complete the populist formula, barely changes how audiences respond. Because engagement data cannot separate approval from amplification, we field an AI-assisted conjoint experiment on nationally representative samples in the United States, Germany, and Poland (N = 6,750). A large language model generates thousands of realistic social media posts in which anti-elite and people-centric appeals vary independently, joining naturalistic stimuli to full experimental control. The experiments show consistently negative effects. Anti-elitism lowers candidate evaluations and vote intention in all three countries. People-centric framing softens the damage but does not undo it, and the two components produce no synergy when combined. We also find no audience that responds positively: respondents who might be expected to welcome the rhetoric—because of their populist attitudes, ideology, personality, or distrust of institutions—punish it less than everyone else but do not seem to favor it. We conclude that populist rhetoric’s influence, wherever it lies, does not lie in mass persuasion.

2026