Here’s a startling revelation: AI chatbots are not just changing political opinions—they’re doing it with alarming effectiveness, even when they rely on inaccurate information. But here’s where it gets controversial: a groundbreaking study published in Science reveals that these chatbots are most persuasive when they flood users with detailed, yet often flawed, data. This raises a critical question: Are we trading truth for influence in the digital age?*
The research, led by Kobi Hackenburg of the University of Oxford, involved nearly 77,000 participants who interacted with AI chatbots from OpenAI, Meta, and xAI. These chatbots were tasked with shifting participants’ views on polarizing topics like taxes and immigration. Shockingly, the AI succeeded not by appealing to morality or personalizing arguments, but by overwhelming users with vast amounts of information—much of which was inaccurate. And this is the part most people miss: the more persuasive the chatbot, the less accurate its claims tended to be. For instance, OpenAI’s GPT-4.5, a newer model, produced significantly less accurate information than its predecessors.
This isn’t just an academic concern. With AI tools like ChatGPT and Google Gemini becoming ubiquitous—44% of U.S. adults use them regularly—the implications are profound. Imagine a world where political campaigns, foreign governments, or even malicious actors exploit this technology to sway public opinion with falsehoods. The study warns that in extreme cases, highly persuasive AI could fuel political unrest or spread radical ideologies.
But it’s not all doom and gloom. Some experts, like David Broockman from UC Berkeley, argue that the study debunks the idea that AI will ‘brainwash’ us. Instead, it suggests that what truly persuades us is the sheer volume of detailed information—a reminder that human curiosity and critical thinking still play a central role. Here’s the kicker: If both sides of a debate use AI, the flood of arguments might cancel each other out, leaving us better informed.
Still, the ethical dilemmas are undeniable. Should we regulate AI chatbots to ensure accuracy? Or is this the price of innovation? The study, funded by the British government and conducted by researchers from Oxford, MIT, and Stanford, is a call to action. As Helen Margetts, a co-author, puts it, we need to understand how AI impacts democracy—before it’s too late.
What do you think? Is the persuasive power of AI a threat to truth, or a tool for enlightenment? Share your thoughts in the comments—let’s spark a debate that matters.
