Three times more persuasive than humans: Artificial intelligence outperforms professional fundraisers.
A recent study by researchers from Oxford University and other institutions found that Anthropic's AI model, "Claude Opus 4.6," outperformed professional fundraisers in persuading people to donate, raising new questions about the growing influence of artificial intelligence. The researchers compared commercial AI models with experienced human fundraisers working for the charity Save the Children.In over 1,000 debates, Claude Opus 4.6 was nearly three times more effective at persuading participants to donate a portion of their student stipends and raised an average of 13% more donations than human professionals, according to a Washington Post report reviewed by Al Arabiya Business.
These findings come from a preliminary research paper that has not yet undergone peer review. The experiment also examined debate performance, with Claude and other advanced models outperforming elite professional debaters by 4.6 percentage points.
But this advantage largely disappeared when researchers restricted AI systems to a word count closer to that of humans. This suggests that the ability to be verbose and present large amounts of information quickly played a greater role in the AI's success than any actual superior reasoning ability.
The researchers noted that chatbots often produced messages much longer than those written by professionals, containing factual claims and expert references. They cautioned, however, that persuasiveness is not necessarily linked to accuracy, explaining that some models presented convincing information that was either unsupported or fabricated.
Interestingly, the researchers themselves caution against overreacting. The experiments relied entirely on written conversations, with participants willing to engage in lengthy discussions of 15 to 20 minutes, which may not accurately reflect real-world behavior. The study also did not test scenarios of human-AI collaboration, which is arguably a more realistic model for future operations.
Nevertheless, the findings highlight a growing reality: AI models are becoming remarkably adept at persuasion. If they can convince people to donate more effectively today than trained professionals, they may tomorrow be just as effective at influencing purchasing decisions, political opinions, or public discourse.
.webp)
0 Comments