AI Chatbots Adopting Ad-Based Model That Polarized Social Media

The introduction of advertising to AI chatbots marks a troubling turning point that could amplify the polarization effects seen in social media, according to researchers at Northwestern University.

OpenAI began testing advertisements in ChatGPT in January 2026, and by February ads were live for hundreds of millions of free-tier users. Within six weeks, the company's ad revenue surpassed $100 million annualized, with more than 600 advertisers participating. Google has announced plans to bring ads to Gemini this year, while Meta already uses AI assistant interaction data to refine ad targeting across its platforms.

The shift represents a fundamental change in how AI systems will be optimized—one that Northwestern professors William J. Brady and Eli J. Finkel argue will repeat the mistakes that transformed social media from a connector into a polarizer.

The engagement trap

Facebook introduced its first engagement-based algorithm in 2009, replacing chronological feeds with popularity-sorted content. Twitter and YouTube followed similar paths, sacrificing the potential to connect people across divides for a business model requiring maximum user attention to generate advertising revenue.

The platforms didn't set out to polarize America—they intended to maximize engagement, and polarization emerged as a byproduct. New research has demonstrated that content associated with intergroup conflict and moral outrage is systematically amplified by engagement-based algorithms because divisive content captures attention and drives the metrics that generate ad revenue.

These algorithms don't merely show users polarizing content; they train users to produce more of it. When posts containing moral outrage receive likes and shares, users learn to create more outrage in subsequent posts.

AI's sycophancy problem

While some studies show chatbots can reduce conspiracy theory beliefs and shift partisan attitudes through evidence-based challenges, those findings describe systems deliberately designed to counter user views. An engagement-optimized system would operate in the opposite direction.

Even current AI chatbots exhibit what researchers call "sycophancy"—the tendency to tell users what they want to hear rather than what is true. Multiple studies have documented this pattern across leading AI assistants, including in politically charged contexts.

OpenAI has stated that "ads do not influence the answers ChatGPT gives you." Facebook made similar assurances about its News Feed, claiming it simply showed users the most "relevant" content. Both then and now, the reassurance misses the point.

Why it matters

The issue isn't whether advertisers can directly alter chatbot responses, but how a business model dependent on user engagement subtly bends design decisions toward maximizing time on platform. Layering advertising incentives on top of existing sycophancy creates a system primed to flatter users—producing prolonged sessions where users luxuriate in validation of preexisting beliefs while becoming further convinced of opponents' foolishness.

Chatbots speak in first person, adapt to reactions in real time, and are trusted by users in ways social media feeds never were. A technology this persuasive, paired with incentive structures rewarding flattery, isn't a corrective to social media's problems—it's a more potent version of them.

Social media companies saw mounting evidence that optimizing for engagement fueled misinformation and eroded democratic health. They ignored it and pressed forward. The consequences manifested in rising sectarianism and fraying social fabric.

The details were first reported in a commentary published by the Chicago Tribune, authored by Brady, an assistant professor of management and organizations at Northwestern's Kellogg School of Management, and Finkel, a professor of psychology and management at Northwestern.