Building advertising for the way people use AI

OpenAI says ChatGPT now reaches 1.2 billion people a week, and it is leaning harder into advertising to pay for that reach. The new piece is a visual ad format: image-based ads showing product inspiration or usage, tested first during image generation in ChatGPT. OpenAI says the ads will be clearly labeled, kept separate from the image being created, and will not influence ChatGPT’s answers. US testing starts later this month with an initial group of advertisers.

The bigger shift is plumbing. OpenAI is connecting ChatGPT Ads to the tools marketers already use — Hightouch, Tealium and LiveRamp for sending conversion data, attribution partners like AppsFlyer, Adjust, Branch, Triple Whale and Northbeam, and geo-based incrementality experiments with Haus, Measured and WorkMagic. It is also piloting brand-suitability evaluations with DoubleVerify and Integral Ad Science that, per OpenAI, run without access to private conversations.

When a platform hands advertisers conversion pipes and third-party verification, it is no longer experimenting with ads. It is building an ad business meant to scale.

We’ll say the uncomfortable part first: ads were always coming. Free access to a frontier assistant for over a billion people costs real money, and OpenAI is candid that advertising lets it serve a much larger market than subscriptions alone. We think broad access to useful AI is a good thing. If ads are the price of keeping a capable free tier, that trade can be defensible.

But the line OpenAI keeps repeating — advertising does not influence the answers — now carries enormous weight, and it is a promise users cannot verify. Search ads at least sit visibly beside a list of results. In a chat assistant, the answer is the product. The moment people suspect a recommendation was shaped by who paid, the whole value of asking an assistant collapses into a sales funnel. We’re glad OpenAI says its placement guardrails keep ads out of emotionally vulnerable or sensitive conversations, and that DV and IAS will test brand safety without reading real chats. Those audits are aimed at advertisers’ comfort, though, not users’. I’m watching for an equivalent independent check that ad money never reaches the answer layer.

The measurement news deserves more scrutiny than the visual format. Partners will send conversion data from advertisers’ systems into ChatGPT Ads. That is standard for the industry, and OpenAI says conversations stay private. Still, this is how every ad platform starts: first the pixels and the APIs, then the pressure to target better, then the slow creep of what counts as acceptable signal. The early results OpenAI highlights are partner-reported and attractive — DV Rockerbox put WeightWatchers’ attributed cost per acquisition 15.3% below its blended paid-search benchmark, and Triple Whale said 93% of Portland Leather’s visitors from ChatGPT Ads were new. Strong numbers like that are exactly what pull budgets away from search, which means the incentive to optimize ChatGPT for advertisers will only grow.

Placing the first visual ads inside image generation is also telling. That is a creative, open-ended moment — someone dreaming up a room, an outfit, a trip. It is the perfect slot to show a product fitting into someone’s life. It is also the moment when people are most suggestible and least on guard.

Our read: this can work, and it may fund a lot of genuinely useful free AI. If it scales without hard, verifiable walls between ads and answers, the assistant people treat as a neutral advisor turns into the most persuasive salesperson ever built.

OpenAI first introduced ChatGPT Ads under a set of published ads principles: answers stay independent, conversations stay private, users stay in control. Today’s update adds a visual format, a long list of measurement and attribution integrations, Negative Phrases for qualifying advertisers, and brand-suitability pilots with DoubleVerify and Integral Ad Science. Sign-ups run through ads.openai.com.