AI Chatbots Outperform Humans in Fundraising and Debating, Study Finds

A groundbreaking study conducted by researchers from the University of Oxford and the AI Security Institute has demonstrated the surprising persuasive capabilities of advanced AI chatbots. The research pitted commercial AI models, including Claude Opus 4.6 and Google's Gemini 2.5 Pro, against human professionals in real-world scenarios, specifically fundraising and competitive debating. In a compelling fundraising experiment, the AI model Claude Opus 4.6 proved remarkably effective. Over more than 1,000 conversations with potential donors for the nonprofit Save the Children, Claude was nearly three times more successful at securing donations compared to a group of experienced human fundraisers. Beyond just a higher conversion rate, the AI also managed to elicit an average donation that was 13% higher than those secured by its human counterparts. The study further explored the chatbots' persuasive power in a separate test involving elite competitive debaters. Here too, AI models like Claude and Gemini 2.5 Pro outperformed the human debaters by a margin of 4.6 percentage points. This indicates a significant ability for AI to construct and deliver compelling arguments. However, the researchers identified a crucial caveat to the AI's superior performance. When the chatbots were subjected to constraints mimicking human communication, such as limiting their response speed and word count per conversation, their persuasive advantage largely vanished. This suggests that the AI's effectiveness is not necessarily due to deeper understanding or emotional intelligence, but rather its capacity to rapidly generate and present a high volume of information and claims. The study also touched upon the factual accuracy of the chatbots, noting that while some models, like OpenAI's GPT 5.4, scored high on truthfulness (89%), others, like xAI's Grok, scored significantly lower (26%). Interestingly, greater truthfulness did not directly correlate with higher persuasiveness in the aggregate. The implications of these findings are profound, particularly for fields like political communications and marketing, highlighting both the potential and the ethical considerations of deploying highly persuasive AI.