OpenAI will start showing picture ads in ChatGPT while you generate images

On Monday, Oct. 5, 2026, OpenAI said it is adding a visual ad format to ChatGPT that uses images to show products and services. It will first be tested in the US later in October, appearing while people generate images, with ads clearly labeled and kept separate from the image being made.

Image generation is one of the main reasons people open ChatGPT, and now it's becoming ad space. When you ask for a picture of a dinner party, a grocery brand can sit right next to it. OpenAI is also lining up the same measurement and brand-safety companies that run ads on the rest of the web, which tells you it wants ChatGPT treated like a full ad network, not an experiment. The test is small for now, but the direction is clear.

On Monday, 5 October 2026, OpenAI published “Building advertising for the way people use AI.” The post says the company is introducing a new visual ad format in ChatGPT, expanding its measurement tools and partnerships, and developing new ways to understand brand suitability. A visual ad, here, is a paid placement built from pictures, not only a line of text. Those lines are OpenAI’s.

What the pictures are for. OpenAI says the format helps people imagine how a product or a service could fit into their lives. The images can show product inspiration, how a product is used, or the experiences a product makes possible. A person might discover a brand, see what makes an offer different, and decide what to buy. For advertisers, OpenAI says, the format is a way to tell that story with pictures.

Where the test starts. OpenAI says it will first try the format while someone is generating an image in ChatGPT. Image generation is the step where ChatGPT draws a picture from a request. The company says the ads will be clearly labeled and will stay separate from the image being created. It also repeats a rule it states for its other ad formats: advertising does not influence the answers ChatGPT provides. Testing begins later in October in the United States, with an initial group of advertisers. The post does not name that group.

How big OpenAI says the audience is. The post says ChatGPT reaches 1.2 billion people each week, and it calls ChatGPT the world’s largest AI-native consumer platform. AI-native, in that phrase, means a product built around a chat model, rather than a website that added a chatbot later. OpenAI says advertising lets it offer AI to more people than subscriptions alone, and gives businesses a way to reach people while they explore ideas, make decisions, and look for products and services. Those sentences are OpenAI’s. The post does not print a price for an ad or a revenue figure.

How an advertiser sends results back. OpenAI says new integrations with Hightouch, Tealium, and LiveRamp make it easier to send conversion data from systems a business already uses into ChatGPT Ads. A conversion, here, is a result the advertiser wanted, such as a purchase or a signup. ChatGPT Ads is OpenAI’s name for the paid placements inside ChatGPT.

Who can assign credit for a later visit or sale. OpenAI says it now supports attribution partners across the web and in apps, including AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin. Attribution means deciding which ad gets credit when someone later visits or buys. The same post says OpenAI also works with full-funnel measurement partners, including Fospha, Measured, and INCRMNTAL. Full-funnel, here, means following a person from the first look at an ad through a later purchase, not only counting the click.

Whether the ads cause extra sales. OpenAI says that work is still early. It is partnering with Haus, Measured, and WorkMagic on geo-based experiments. Geo-based means comparing places where the ads run with places where they do not, to see whether the ads cause extra sales. OpenAI calls that the causal impact of ChatGPT Ads. Causal, in that sentence, means the ad changed the outcome. A sale that would have happened anyway does not count as that impact. Those experiments are OpenAI’s plan, and the post says they are in early stages.

What three measurement partners reported, as OpenAI cites them. These figures are the partners’ reports. They are a citation on OpenAI’s post, and they are not a separate audit of ChatGPT’s ad system. DV Rockerbox said WeightWatchers’ attributed cost per acquisition on ChatGPT Ads was 15.3% lower than its blended paid-search benchmark. Cost per acquisition is what that report counts as the spend to gain one new customer. A blended paid-search benchmark is WeightWatchers’ usual cost across paid search, the ads that appear in search results, rolled into one comparison number. WorkMagic said a wellness brand called Dose saw a statistically significant lift, and that 67% of the extra purchases came from customers who were new to the brand. Statistically significant, in that report, means the partner judged the increase unlikely to be chance. Triple Whale said 93% of Portland Leather’s visitors who arrived from ChatGPT Ads were new. Those three numbers are the partners’, as OpenAI prints them.

Where ads are allowed to sit, and who does not see the chat. OpenAI says its placement rules are meant to keep ads out of conversations that are emotionally vulnerable, sensitive, or otherwise a poor fit. It is piloting brand-suitability checks with DoubleVerify and Integral Ad Science. Brand suitability means whether the setting around an ad matches what an advertiser is willing to appear next to. OpenAI says those partners will look at how its rules are applied in a controlled test, and that the check will not give them access to private user conversations. A control called Negative Phrases is now available to qualifying advertisers, so a brand can name phrases it does not want its ads near. Those lines are OpenAI’s.

Where a business signs up. OpenAI says businesses can sign up for ads at ads.openai.com. The post also points to a longer note on the measurement tools at ads.openai.com/blog/more-ways-to-measure. It does not print ad prices, a revenue figure, a launch date outside the United States, or the names of the advertisers in the first test.

The picture is OpenAI’s own example of the new format. Two phones sit side by side on a dark blue field. On the left, someone has asked for a photo of a dinner party. The reply says it is putting that together, and a picture of people around a dinner table is on the screen. Under that picture is a labeled ad: Heirloom Grocery, marked Sponsored, with the line “From dinner party to grocery list” and four grocery photos. The box at the bottom reads Ask ChatGPT. The phone on the right shows the same kind of screen, with a different dinner photo and the same grocery ad. Heirloom Grocery is the name on this example. The post does not list it among advertisers in the US test. The frame does not print a calendar date.

In plain terms, OpenAI said on Monday that ChatGPT will test picture ads in the United States later this month, shown while a person is generating an image. The company says those ads will be labeled and kept apart from the picture being made, and that ads do not change ChatGPT’s answers. It is also connecting ChatGPT Ads to measurement and brand-safety firms that already score ads on the rest of the web, and it is citing early results those partners reported for WeightWatchers, Dose, and Portland Leather. The first test group is unnamed, and the post does not print a price.