ChatGPT Ran 500 Ad Impressions Against a Reporter. Here's What OpenAI Is Selling.

OpenAI's decision to introduce advertising to ChatGPT's free tier was, at the time of announcement, treated primarily as a financial story: the company needed revenue beyond subscriptions, and ads were the obvious answer. What was less examined was the mechanics of how those ads would work — what signals they would use, how they would relate to conversational context, and what the experience would actually feel like for users.

Wired's Katie Notopoulos spent several weeks asking ChatGPT 500 questions, logging every ad impression she received. The resulting analysis offers the most detailed public map of OpenAI's ad system available.

What the Ads Are and Are Not

Based on the sample, ChatGPT's ads are not simple display banners or contextual keyword matches in the traditional sense. They appear to use a more sophisticated targeting layer — one that responds not just to keywords in a query but to inferred intent behind it. A question about productivity tools surfaced ads for enterprise software. A question about learning a new language surfaced language-learning platform ads. A question about joint pain surfaced pharmaceutical advertising.

The targeting appears to be probabilistic and intent-driven rather than keyword-matched. This is a meaningfully different ad product from what Google Search has historically offered. Google matches ads to query terms; ChatGPT appears to match ads to inferred user objectives. The distinction matters for advertisers (higher conversion probability) and for users (more accurate targeting of their needs, but also more exposure of their intent to commercial systems).

The Super Bowl Context

OpenAI's ad launch arrived in awkward proximity to Anthropic's Super Bowl campaign, which explicitly positioned Claude as the ad-free alternative to ChatGPT. That framing proved effective: Claude's daily downloads surpassed ChatGPT's in the weeks following the Super Bowl. OpenAI's ad rollout may have validated the Anthropic message more than it advanced OpenAI's revenue goals.

The underlying tension is structural. OpenAI's free tier serves hundreds of millions of users; monetizing that base without subscription revenue requires either ads or data licensing. Ads were the faster path. But in a market where the premium alternative is explicitly positioning itself as free from advertising, every ChatGPT ad impression is also an implicit argument for switching to Claude.

What This Signals About AI Business Models

The ChatGPT ad rollout is one of the first large-scale tests of advertising inside a conversational AI product at consumer scale. The results will have industry-wide implications. If the model proves successful — high CPMs, strong conversion, manageable user attrition — every AI company with a free tier will face pressure to follow. If users respond by upgrading to paid tiers or switching to competitors, the economics will push the industry toward subscription purity.

Either outcome tells the market something important. OpenAI is, in effect, running a public experiment on whether advertising and AI conversations can coexist at scale without destroying the product's core value proposition. The 500-question sample suggests users are tolerating the ads, at least initially. Whether tolerance holds over millions of interactions and months of use is the question the experiment has yet to answer.