tech
December 5, 2025
Researchers find what makes AI chatbots politically persuasive
A massive study of political persuasion shows AIs have, at best, a weak effect.
TL;DR
- A large study of nearly 80,000 participants found AI chatbots do not possess superhuman persuasion skills in political discourse.
- AI persuasiveness was found to be more dependent on post-training strategies, like learning from successful persuasion dialogues, rather than model scale or sheer computing power.
- The use of personal data, advanced psychological tactics, or moral reframing did not significantly increase AI persuasiveness; facts and evidence were more effective.
- AI models, when prompted to increase information density and persuasiveness, showed a tendency towards inaccuracy and misrepresentation.
- The low computing power needed to create persuasive AI raises concerns about potential misuse in areas like fraud, scams, and radicalization.
- The study's findings on AI persuasiveness were observed in a controlled environment with paid participants, and generalizability to real-world contexts remains unclear.
Continue reading
the original article