A study conducted by researchers including those at the University of Oxford reveals that the AI chatbot Claude Opus 4.6 significantly outperforms human fundraisers in persuading people to donate money, sparking concerns about the growing power of AI-driven communication.
- Claude AI raised more frequent and larger donations than human fundraisers
- AI outperformed elite debaters but mainly due to verbosity and info volume
- Researchers caution about AI accuracy and real-world application limits
What happened
Researchers from the University of Oxford and other institutions conducted a study comparing Anthropic’s AI model Claude Opus 4.6 against professional fundraisers working for Save the Children. Over more than 1,000 participant conversations, Claude proved nearly three times as effective at persuading individuals to donate money and generated donations averaging 13% larger than those raised by experienced human fundraisers.
Additionally, the study evaluated debate performance, finding Claude and other state-of-the-art AI systems outperformed top human debaters by 4.6 percentage points. However, this advantage disappeared when AI output was restricted to the same word count as humans, indicating that AI’s length and information density contributed heavily to its edge.
Why it matters
These findings highlight how advanced AI models are rapidly improving at nuanced social tasks that involve persuasion, raising important ethical and societal questions. With AI outperforming trained fundraisers, similar techniques could soon impact consumer behavior, political opinions, and public discourse, making transparency and regulation more critical than ever.
However, researchers emphasize that persuasive AI communication does not guarantee accuracy. The models sometimes produced convincing but inaccurate or unsupported claims, underscoring risks of misinformation. Moreover, the artificial, controlled environment of the study—with participants willing to engage in lengthy chat sessions—may not fully reflect real-world human interactions.
What to watch next
Future research will likely explore how AI-human collaboration in persuasion tasks might amplify or moderate these effects, as real-world scenarios are expected to blend automated and human efforts rather than rely solely on AI. Monitoring developments in AI communication transparency and ethical guardrails will be essential as these models become more integrated into everyday applications.
Regulators and industry leaders should also focus on establishing frameworks to ensure AI-generated content is both clearly identified and held to standards that prevent misuse. How society balances AI’s productivity benefits against potential risks from inaccurate or manipulative messaging will be a key area to watch.