OpenAI will start testing a new visual ad format in ChatGPT in the US later this month, the company said. The ads will first show up when users generate images. It's the first expansion of ChatGPT's ad business since OpenAI brought ads to the chatbot in February.
For now, the format stays separate from whatever the model produces. Ads won't influence ChatGPT's answers, and they won't appear at all for people on Plus, Pro, or any other paid tier.
From a name and a link to something you look at
Until this point, ChatGPT's ad inventory has been pretty bare-bones. A company's name, its logo, and a link to its product, tucked into a sponsored section below the chat. That's it. Static, text-heavy, easy to scroll past.
📊 Market Data Snapshot
A visual format tied to image generation is a different animal. Image prompts are inherently visual, and a static banner next to a rendered picture is a more natural place to sell something than a line of grey text under a wall of prose. It's also the kind of placement that advertisers have been asking for since the earliest days of search advertising — something that sits next to the output, not buried under it.
The obvious question is whether users will tolerate it. OpenAI is drawing a clear line: the ad is separate from the generated image. That's a sensible boundary to plant early, because the moment an ad starts bleeding into a model's output, the trust problem gets much harder to walk back.
Why image generation first
Image generation is one of the most compute-intensive things ChatGPT does, and it's also one of the stickiest features for casual users. Serving a visual ad next to an image gives OpenAI a format that matches the medium. It also gives the company a way to monetize a feature that, until now, has only cost it money to run.
Pushing this out in the US first is the low-drama option. Regulatory scrutiny of AI advertising is lighter at home than in Europe, and the US user base is large enough to generate real signal on click-through and engagement before any international rollout. OpenAI hasn't said when — or whether — the test extends beyond the US.
Free users pay in attention, paid users don't
The tier split is the part worth watching. Subscribers to Plus, Pro, and other paid plans won't see the visual ads at all. That keeps the paid pitch clean: pay, and the ads go away. It's the same bargain streaming services made for years before they started chipping away at it.
Whether OpenAI holds that line is a separate matter. Ad revenue scales with impressions, and free-tier users are the ones generating impressions. The company has given no indication it plans to put ads in front of paying customers, but the economics of AI inference aren't getting cheaper.
What isn't being said
OpenAI hasn't named advertisers, shared pricing, or given a firm launch date beyond "later this month." It also hasn't said how it will label the ads inside the interface, or whether image-generation ads will be sold programmatically or through direct deals.
Those details matter more than the format itself. If the test runs through the end of the year without a privacy backlash, expect OpenAI to extend visual ads to other parts of the product. If it doesn't, the sponsored section stays what it's been since February — a small line of text under a chatbot's answer.
