OpenAI announced today that a new visual ad format will begin testing in the US later in October 2026, placing sponsored imagery directly alongside the results of ChatGPT image generation. The pitch is that an ad arriving mid-prompt catches a user already imagining a room, an outfit, or a party, which is to say a user already halfway into a purchase.
The format sits beside the generated image rather than inside it. OpenAI says ads will be clearly labeled, won’t influence ChatGPT’s answers, and will stay out of sensitive conversations like personal health. Plus, Pro, and Enterprise subscribers won’t see them at all.
What’s more consequential than the pixels is the plumbing. OpenAI is wiring the format into Hightouch, Tealium, and LiveRamp for conversion data, and TechCrunch counts ten attribution partners at launch: AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin. Fospha, Measured, and INCRMNTAL handle full-funnel measurement, with Haus, Measured, and WorkMagic running early-stage geo-based incrementality tests. That’s the same stack a mid-sized Shopify operator already runs. The learning curve is near zero.
OpenAI paired the announcement with three partner numbers. DV Rockerbox reported WeightWatchers’ attributed CPA on ChatGPT Ads was 15.3% below its blended paid-search benchmark. WorkMagic clocked statistically significant lift for Dose, with 67% of incremental purchases from net-new customers. Triple Whale found that 93% of Portland Leather’s ChatGPT Ads visitors were new. Early-partner metrics are early-partner metrics, but the shape of the data (net-new reach, competitive CPA) is the shape OpenAI needs to show.
There’s no stated minimum spend, a conspicuous contrast with ChatGPT’s CRM-targeting rollout and its $25,000 user floor earlier this month. Combined with sponsored agents that turn ChatGPT ads into a sales conversation, OpenAI is now shipping three distinct ad surfaces in roughly a week.
The constraint is inventory. ChatGPT reaches 1.2 billion people weekly and runs about 2.5 billion daily prompts globally against Google’s 15 billion searches, with only one ad placement per response. Zeno Group’s Shamsul Chowdhury put it plainly to Digiday: “which all leads to a disconnect between supply and demand.” Launch access is limited to what OpenAI calls an “initial group of advertisers,” with no public application path disclosed.
That combination (open attribution, closed roster) is the real design choice here. OpenAI is building the measurement rails that make the format legible to anyone, while keeping the on-ramp narrow enough to control the first wave of case studies. The accessibility is architectural. The access is editorial.
Sources
- https://openai.com/index/new-chatgpt-ads-format-and-measurement/
- https://techcrunch.com/2026/10/05/openai-launches-visual-ads-that-appear-alongside-image-generation-results/
- https://digiday.com/marketing/openai-moves-to-make-chatgpt-ads-more-measurable-and-more-visual/
- https://www.engadget.com/2277271/openai-is-testing-more-ads-in-chatgpt/
- https://www.bleepingcomputer.com/news/artificial-intelligence/openai-will-show-visual-ads-in-chatgpt-while-you-generate-images/