ChatGPT Is Now Showing Ads While Images Are Being Generated

Waiting for ChatGPT to finish an image just became advertising real estate. Users on the ad-supported tiers are beginning to see sponsored visuals appear while their image renders, a format OpenAI said last month it would test.

X user Juozas Kaziukėnas posted a screenshot on X on Tuesday of what he was served. A progress indicator sat at 51%, with the line “Here’s an ad while we create your image” beneath it. Below that was a labeled ad from Blue Buffalo, a pet food brand, with a puppy photo, a headline about “smart nutrition for growing puppies,” a short line on calcium and DHA, and a “Learn more” button. Kaziukėnas noted that the ad format was announced about a month ago, that he started seeing it today, and that his conversation had nothing to do with dogs or pet food.

What OpenAI has said about the format

OpenAI announced the visual ad format on October 5, saying it would begin testing during image generation in the US with an initial group of advertisers. The company said the ads would be clearly labeled and kept separate from the image being created, and it repeated its standing line that ads do not influence ChatGPT’s answers. In its announcement, OpenAI described the format as a way for brands to show product inspiration, how a product is used, or the experience a service offers, and noted that ChatGPT reaches 1.2 billion people a week.

The test builds on ads that have been running since February 9, when OpenAI began showing sponsored placements below ChatGPT’s responses in the US. OpenAI’s help center page on ads now lists the image-generation unit alongside those: a multi-image visual ad that is labeled and shown separately from both ChatGPT’s response and the image being created.

The ground rules are the same as before. Ads appear only on the Free and Go plans, while Plus, Pro, Business, Enterprise and Edu accounts stay ad-free. They aren’t shown to accounts identified as belonging to under-18 users or in Temporary Chats, and they aren’t eligible to appear near sensitive topics such as personal health, mental health and politics. Political advertising isn’t allowed, and advertisers don’t receive chats, chat history, memories or personal details. Free users can opt for an Ads-Free experience, but it comes with fewer messages per day and no access to tools like image generation.

That last tradeoff is worth noticing. The image tool is now both something free users can use only if they accept ads and the place where the newest ads are being tested.

How are the ads selected?

OpenAI says ads are chosen by matching topics from the current chat thread to ads submitted by advertisers, with additional signals used if ad personalization is on. A pet food ad against an unrelated image request doesn’t fit that description. OpenAI hasn’t said how matching works for the new image format, so it’s unclear whether this was a broad placement, a gap in an early test, or something else. It’s a single data point from a test that OpenAI has framed as limited, and it will be interesting to see whether relevance tightens as the format reaches more users.

A new kind of ad slot

The slot itself is genuinely new. Image generation takes seconds to a minute, and the person is already committed to waiting for the result. Earlier digital ad formats had to interrupt something (a video, an article) or compete with a list of search results. A sponsored visual that appears during that dead time borrows attention that was otherwise empty, and a visual format suits the task, since the user is already looking at a screen waiting for a picture.

It also hints at where AI-native advertising could go. Someone generating a living-room mockup is expressing a visual intent, and furniture or paint brands can show products in that moment, which is what OpenAI’s own description of the format (product inspiration, how products fit into life) points toward. Ads can also become conversations: OpenAI’s help page says users can tap the three-dot menu on an ad and select Ask ChatGPT, which shares the ad with the chatbot so it can answer questions about it, turning a static message into something closer to talking with a salesperson. And conversational products are becoming places where purchases happen, with OpenTable’s native ChatGPT integration driving 190,000 seated diners in September, according to its parent company, so sponsored placements sit naturally beside that kind of intent.

That commerce angle also suggests where the money may land first. Autonomous agents that book things for people haven’t taken off: Uber said agents accounted for only about 500 Eats orders and four rides in Halloween week, and Airbnb called its agent bookings non-material. Putting brands in front of people inside the conversations they’re already having is a much easier thing to build and sell.

Why ads are what keeps AI free

It’s easy to groan at an ad appearing while you wait, but the economics explain why OpenAI is doing it. Industry estimates put paying subscribers at roughly 5% of ChatGPT’s weekly users, which leaves the vast majority of people using a compute-heavy product for nothing. OpenAI says keeping the Free and Go plans fast and reliable requires significant infrastructure and ongoing investment. Image generation is among the more GPU-hungry things a free user can ask for.

Ads are how the internet has long solved this problem. Search, maps, email and video are free to billions of people because advertisers pay for the attention. The same logic applies to AI: a student, a small shop owner or a first-time user who can’t justify a monthly subscription gets access to a capable model, and the cost is borne by brands that want to reach them. OpenAI has shown the model works at least commercially. Its ad pilot passed $100 million in annualized revenue in under two months.

CEO Sam Altman has come a long way on this. In 2024 he called ads a “last resort”, and when ads were announced in January he said, in essence, that many people want to use lots of AI and don’t want to pay, so he hoped an ad model could work. That is the democratization argument in OpenAI’s own words: if access depends entirely on subscriptions, a large part of the world is priced out of the most useful information tool of the decade.

Posted in AI