OpenAI is sticking more ads in ChatGPT

OpenAI announced Monday that it will begin testing visual ads in ChatGPT in the US later this month, The Verge’s Emma Roth reports. The new format puts images of sponsored products and services on screen and will initially appear when people generate images. Until now, ChatGPT ads — which first arrived in February — were limited to a company’s name, logo and a link in a sponsored section beneath the chat.

OpenAI says the visual ads stay separate from the image you create and do not influence ChatGPT’s answers. Plus, Pro and Enterprise subscribers won’t see them. The company says it is testing with an initial group of advertisers and plans more formats later. That last line is the one we’d underline: this is a roadmap, not a one-off.

We’re not going to pretend ads are scandalous on their own. Running frontier models for free users costs real money, and ads are how a lot of the internet keeps a free tier alive. If sponsored images are what keeps capable image generation available to people who can’t pay a monthly subscription, that’s a trade many users will accept.

The problem is what ads do to a product over time. Image generation is an obvious first venue — someone making a picture of a living room is one step from a furniture ad. But the commercial pull doesn’t stop at the easy cases. Every ad business eventually wants more placement, more targeting and more formats, and OpenAI has already said more formats are coming. I’m watching whether the line between a sponsored panel and the assistant’s own suggestions stays bright once revenue targets get set.

OpenAI’s two promises carry the weight here. First, that ads won’t influence answers. That’s the right commitment and the hardest one for outsiders to verify; we’d like to see independent audits, not just assurances. Second, that guardrails keep ads out of contexts it calls emotionally vulnerable or sensitive. People bring health worries, grief and money stress to chatbots. A system that knows that much about a user is an unusually powerful ad surface, which is precisely why the restraint has to be enforceable rather than a policy page.

There’s also a two-tier effect worth naming. Paying subscribers get a clean assistant; free users get the ad-supported one. If this scales, the quality and neutrality of AI help starts to depend on who can afford it.

OpenAI introduced ads in ChatGPT in February with a text-and-logo format in a sponsored section under responses. The visual format is the first expansion of that program, starting with image generation in the US.