Getting AI to write emails or debug code is one thing. Getting it to convince people to part with their money is another matter entirely. According to a new report by The Washington Post, a new study from researchers at the University of Oxford and other institutions found that Anthropic’s Claude Opus 4.6 outperformed professional human fundraisers when it came to persuading people to donate, raising fresh questions about the growing influence of AI.
Claude beat human fundraisers, but there is an important catch
The researchers pitted commercial AI models against experienced fundraisers working on behalf of the charity Save the Children. Across more than 1,000 conversations, Claude Opus 4.6 was nearly three times as effective at convincing participants to donate part of their study bonus and also secured donations that were, on average, 13% larger than those raised by human professionals. The findings come from a preprint paper that has not yet undergone peer review.
The study also explored debate performance, where Claude and other frontier models outperformed elite competitive debaters by 4.6 percentage points. However, the advantage largely disappeared once researchers limited the AI systems to using roughly the same number of words as their human counterparts. That suggests verbosity and the ability to rapidly surface large amounts of information played a major role in the AI’s success, rather than some fundamentally superior reasoning ability.
The researchers noted that the chatbots often produced messages several times longer than those written by professionals and packed them with factual claims and expert references. They also warned that persuasiveness did not necessarily correlate with accuracy, pointing out that some models generated convincing but unsupported or fabricated information.
The worrying part isn’t fundraising. It’s what comes next
Interestingly, the researchers themselves caution against overreacting. The experiments relied entirely on written conversations, with participants willing to engage in lengthy 15 to 20-minute exchanges that may not reflect real-world behavior. They also did not test scenarios where humans and AI collaborate, which is arguably the more likely future workplace model.
Even so, the findings highlight a growing reality: AI models are becoming remarkably good at persuasion. If they can convince people to donate more money than trained professionals today, tomorrow, they could be just as effective at influencing purchasing decisions, political opinions, or public discourse. That’s exciting from a productivity standpoint, but it also underscores why transparency and safeguards around AI-generated communication are becoming more important than ever.