OpenAI announced on October 5 that ChatGPT will show visual ads while users generate images, with testing starting later this month for an initial group of advertisers in the US. It's the first time ChatGPT ads have entered the image-generation flow — previous ad placements all sat inside text conversations.
Where the Ads Appear
According to OpenAI, the new format appears during the image-generation process, clearly labeled as an ad and shown separately from the image the user is generating. Advertisers can use it to showcase product inspiration, usage examples, or a particular service experience. OpenAI hasn't released a sample image, but one scenario might look like this: a user asks ChatGPT to render a Scandinavian-style living room, and a furniture brand's material appears next to the generation panel.
OpenAI reiterated that ads won't affect ChatGPT's answers. The company also said it has built in safeguards so that ads won't appear in conversations that are emotionally vulnerable, sensitive, or otherwise inappropriate; the independent partner responsible for evaluating this safeguard has no access to users' private conversations.
The latest user figure given in the announcement is 1.2 billion weekly active users.
Measuring the Effect First
In the same announcement, OpenAI rounded out a set of measurement tools, with partners split into three categories:
| Use | Partners |
|---|---|
| Conversion tracking | Hightouch, Tealium, LiveRamp |
| Attribution | AppsFlyer, Adjust, Branch, Triple Whale, Kochava |
| Brand-safety verification | DoubleVerify, Integral Ad Science |
Several names in the attribution column specialize in mobile app-install attribution — app developers running ChatGPT ads can use the same measurement standard to compare return on spend against other channels. DoubleVerify and Integral Ad Science are the third-party firms brands commonly use to check what content their ads run alongside.
Image generation is more sensitive when it comes to brand safety. Advertisers have no control over what image a user generates, and an ad creative sitting next to an inappropriate image is far more jarring than one appearing beneath a block of text. Bringing in two verification firms from the start gives large brands a checkpoint they can write into their ad contracts.
Four Steps in Eight Months
Lay out the ChatGPT ads timeline end to end. On February 9, ads were tested on a small scale with US Free- and Go-tier users. On May 5, the self-serve ad management dashboard opened to all US advertisers. On May 7, OpenAI announced ads expanding to the UK, Brazil, Japan, and South Korea, joining the existing US, Canada, Australia, and New Zealand markets, bringing the addressable market to nine countries. By October 5, the ad placement extended from text conversations into image generation.
The first three steps settled who can buy and where; the fourth moved where the ad actually sits. An ad beneath a text reply competes for attention with the answer itself; generating an image takes a moment, and during that wait, the user's eyes stay on the screen — the interface itself is open inventory.
A report on this site in May noted that OpenAI had internally set this year's ad revenue target between $2.4 billion and $2.5 billion. Adding another ad placement is the most direct way to pad that target with more inventory.
What the Announcement Left Out
This announcement didn't mention pricing. When the dashboard opened in May, reports suggested cost-per-click bids ranged from $3 to $5; OpenAI hasn't said whether image ads will keep the click-based pricing or switch to impression-based billing.
There's also no timeline for how long the test will run or when it will expand to the other eight markets beyond the US. Whether paying subscribers will see these ads also wasn't addressed; so far, ChatGPT ads have only shown up for Free- and Go-tier users.
For users in mainland China, ChatGPT isn't officially available there, and this test is US-only, so this ad placement isn't something they'll see anytime soon.
Sources: OpenAI official announcement, BleepingComputer, CocoLoop; scope of testing, measurement partners, and weekly active user figures follow OpenAI's own announcement.