AI chatbots may be the antidote to 15 years of social media-driven polarization - pushing users toward moderate, expert-aligned positions rather than amplifying radical voices, a new FT data analysis suggests.
John Burn-Murdoch, the FT's data journalist, tested the latest versions of the most widely used AI chatbots against 61 topics from the Cooperative Election Study, spanning social values and policy preferences across a wide range of areas, "Hvylya" reports, citing the Financial Times.
The results, Burn-Murdoch wrote, "strongly support the theory of AI chatbots as depolarising and technocratising." All platforms - Grok, GPT, Gemini, and DeepSeek - nudged users away from the most extreme positions and toward more moderate stances, though each platform had its own ideological lean.
The underlying mechanics explain why. Social media companies make money from attention, which in practice means rewarding sensationalism with little regard for truth. AI companies face a fundamentally different set of incentives: they compete to deliver accurate, intelligent tools that customers pay for, and they bear liability when their products surface harmful content.
Burn-Murdoch drew a historical parallel. "Every media revolution has transformed who distributes information, what messages are distributed and what form they take," he wrote. The printing press and TikTok, despite arriving 600 years apart, both widened the pool of publishers and amplified radical voices. Radio and television pushed the opposite way, creating a monopoly for the voices of elites and experts. AI chatbots, Burn-Murdoch argued, fall into the latter category.
The methodology weighted each simulated conversation as 80 percent the user's original position and 20 percent the chatbot's response, in line with experimental evidence on AI persuasion. The result represents an estimate of the impact of population-wide chatbot usage on political attitudes - grounded in real-world data rather than hypothetical scenarios.
Also read: "Hvylya" reported on how OpenAI set a September deadline for its first autonomous AI research intern.