Ever felt like you're great at winning arguments? A new study might make you think twice.

Research published in Nature Human Behaviour on May 19, 2025, reveals that OpenAI's GPT-4 outperforms humans in debates by a significant margin—but only when it's armed with basic personal information about its opponent.

When given just six demographic details (gender, age, ethnicity, education, employment status, and political views), AI was 64% more persuasive than human debaters. Without this information, both performed similarly.

"When provided with even just this very minimal information, GPT-4 was significantly more persuasive than humans," explains researcher Francesco Salvi from EPFL. "It was quite simple stuff that normally can also be found online in social media profiles."

The study involved 900 US participants debating topics ranging from school uniforms to fossil fuel bans.

What makes this finding particularly concerning is how easily such personal data can be obtained online. As we increasingly engage with AI systems, perhaps we should think twice about what information we're sharing—and who (or what) might be using it to change our minds.