AI outperforms elite debaters at persuasion

What happened

Researchers from Oxford, Stanford, the UK AI Security Institute, and the London School of Economics studied how well AI systems persuade humans across 18,978 conversations involving 6,923 people. AI systems—particularly Anthropic's Opus 4.1 and Opus 4.6—outperformed every class of human persuader tested, including elite debaters and professional fundraisers. When AI spoke to real donors, it raised substantially more real-money donations to Save the Children than professional canvassers from a UK fundraising firm, exceeding them by 10.8 percentage points of a £1 bonus.

Why it matters

The ability of AI to out-persuade humans at scale could reshape power dynamics in society. Those who control these systems could consolidate influence, or conversely, persuasive AI made widely available could help under-resourced actors—small charities, grassroots activists, public defenders—compete against better-funded rivals. The research shows that human coaching narrowed but did not close the gap, suggesting that this is not a skill deficit humans can easily overcome. Society now faces a choice about how to monitor and allocate this capability.

What to watch

The researchers found that AI's persuasive edge came largely from its speed and volume of information delivery. When constrained to write human-length messages at human writing speeds, AI's advantage over coached elite debaters collapsed to non-significant levels. The question ahead is no longer whether AI can out-persuade humans, but how, where, and on whose behalf this capability will be exercised.