AI Is More Persuasive Than World-Champion Debaters

Oxford researchers found that frontier AI was more persuasive than laypeople, tournament debaters, and professional canvassers across four experiments involving about 19,000 conversations.

The advantage showed up even in real-money fundraising, and appeared to come mainly from faster, denser information processing and use.

Human performance improved only when the AI was slowed down to human pace.

The study also found a tradeoff: models that were more persuasive tended to be less accurate, raising concerns about politics, lobbying, scams, and misinformation.