On July 22, 2026, OpenAI did something Sam Altman once called “uniquely unsettling.” The company opened “Advertise in ChatGPT” — a self-serve Ads Manager that lets brands buy their way into your conversations with the AI. Best Buy, Lowe’s, and VistaPrint were among the first to sign up. The launch arrives less than two years after Altman told an interviewer that “ads plus AI is sort of uniquely unsettling to me” and described advertising as “a last resort.”
The predictable outrage cycle has already begun. Altman sold out. ChatGPT is enshittifying. Free users were always the product. All of it true, in its way, and all of it missing the point.
The ads themselves are not the story. The targeting is.
What “Conversational Context” Actually Means
Search ads target what you type into a box. Social ads target your profile, your likes, your browsing history. ChatGPT’s ad platform does something different: it reads the conversation you’re having and decides, in real time, what you’re probably in the market for.
OpenAI calls this “conversational context” targeting. Advertisers don’t bid on keywords. They provide “context hints” — descriptions of the kinds of discussions where their ads might feel relevant. The AI then evaluates your ongoing chat, your history, and your previous ad interactions to decide whether to slip a sponsored message into the thread.
This is not a semantic distinction. A keyword ad appears because you typed “plumber near me.” A conversational context ad appears because you mentioned your basement flooded and the model inferred you might need a plumber — even if you were just telling a story about last year. The difference is the difference between answering a question and reading a diary.
A media buyer who’s been testing the platform since the February beta sent me a Slack DM yesterday during a campaign setup. The “context hints” field let them target “users discussing home renovation frustrations” — not people searching for contractors, just people complaining about their kitchens. “It’s like reading someone’s journal and sliding a Lowe’s flyer between the pages,” they wrote. “Except the journal is live, and the flyer looks like part of the conversation.”
The Assistant That Watches You Think
The privacy implications here are not subtle, but they are novel enough that the existing regulatory framework barely touches them. GDPR and CCPA were written for an internet where data collection meant cookies, tracking pixels, and form submissions. They were not written for a system that ingests your half-formed thoughts, your idle curiosities, your late-night what-ifs, and monetizes the inferences in real time.
OpenAI has been careful about the rollout. Ads appear only on the Free and ChatGPT Go tiers. Paid subscribers — Plus, Pro, Team, Enterprise — are exempt. The company has also limited the platform to a handful of markets: the U.S., Canada, the U.K., Australia, New Zealand, Brazil, Japan, South Korea, and Mexico, with capabilities varying by region.
But the tiering strategy is itself revealing. The message is clear: privacy from conversational surveillance is now a premium feature. If you can afford $20 a month, your chats stay ad-free. If you can’t, your AI assistant is also your ad broker. This is not a bug in the business model. It is the business model.
OpenAI has projected $2.5 billion in ad revenue this year, with a target of $100 billion by 2030. Those numbers don’t come from banner ads. They come from a targeting apparatus that knows more about your intentions than any search engine ever could — because you’re not just telling it what you want to find. You’re telling it what you’re thinking about, what you’re worried about, what you’re planning, what you’re confused by. Every one of those signals is a bid request.
The Competitors Are Taking Notes
Here is the part the outrage cycle will miss entirely: the targeting infrastructure OpenAI is building today will not stay inside ChatGPT.
Anthropic has so far resisted advertising. Google has its own AI products and its own ad machinery, but the two have not yet been fully merged in the way OpenAI is attempting. Apple’s approach to AI is built on on-device processing and privacy as a differentiator. But $100 billion revenue targets have a way of concentrating the mind. Every major AI lab is watching this rollout. If conversational context targeting works — if it delivers the kind of return on ad spend that early numbers suggest — the pressure to adopt it will be existential.
U.S. AI search ad spend is projected to grow from 0.7% of all search ad revenue to 13.6% by 2029, according to eMarketer. That shift represents tens of billions of dollars moving from traditional search to AI-mediated conversations. The companies that control the targeting technology will control the revenue. The companies that don’t will be left selling API access while someone else monetizes the user.
The irony is that Altman was right in 2024. Ads plus AI is uniquely unsettling — not because advertising is inherently corrupting, but because AI assistants are uniquely positioned to observe us in ways no previous technology could. A search engine sees your queries. A social network sees your posts. An AI assistant sees your thinking process. Monetizing that last category is not just another ad platform. It is a new category of surveillance, and we are only beginning to understand what it means.
The ads are fine. The mind-reading isn’t.