OpenAI announced on October 5 a visual ad format that’ll appear alongside ChatGPT image-generation results, along with the first set of named third-party CPA benchmarks for its eight-month-old ad business. The U.S.-only test goes live later this month. The announcement reframes ChatGPT Ads from a budget-at-your-own-risk experiment into something a performance marketer can model.
The three disclosed benchmarks do the heavy lifting. DV Rockerbox attributed WeightWatchers a cost per acquisition 15.3% below its blended paid-search baseline. WorkMagic measured a statistically significant lift at Dose, with 67% of incremental purchases coming from net-new customers. Triple Whale found that 93% of Portland Leather’s visitors from the channel were new. For a small business already running Google or Meta, these are the first numbers that let a media planner wire ChatGPT into the same measurement stack and compare the output honestly.
Which is the point. OpenAI’s attribution roster now spans 10 click-attribution partners (AppsFlyer, Adjust, Branch, Northbeam, Singular, Kochava, Airbridge, Tenjin, Triple Whale, and DV Rockerbox), three full-funnel vendors (Fospha, Measured, INCRMNTAL), and three geo-based incrementality shops (Haus, Measured, WorkMagic). OpenAI concedes the geo work is still exploratory. The click-attribution tier isn’t.
The visual format itself is adjacent to the generated image, not embedded in it. OpenAI says the units will be clearly labeled, kept separate from the generated image, and won’t influence ChatGPT’s answers. That’s the editorial firewall every search-era platform eventually negotiates with its advertisers, and it never holds in the shape it launches in.
The structural constraint is scale, not quality. “There are about 2.5 billion prompts globally happening each day on ChatGPT, whereas there’s about 15 billion searches on Google so the volume is inherently smaller,” Shamsul Chowdhury, EVP paid social at Zeno Group, told Digiday, adding that “there’s only one ad placement on ChatGPT whereas Google has multiple.” ChatGPT’s reach of 1.2 billion people a week sounds enormous until you set it against the inventory depth of mature search. Digiday has reported all year that measurement gaps kept budgets at test level since the February 2026 launch; those gaps just narrowed, but the supply-demand mismatch Chowdhury flags caps near-term scale regardless of CPA.
For owner-operators of image-led brands (home goods, apparel, food, services), the creative surface is finally interesting. The practical gating remains elsewhere: CRM targeting still sits behind a $25,000-user audience floor that excludes most small advertisers, and the ad surface keeps expanding into sponsored agents inside ChatGPT conversations that let a shopper interview a brand before clicking. No self-serve entry point for the visual unit has been disclosed.
The number OpenAI most wants in circulation is 15.3%. That one is now in the record.
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://thenextweb.com/news/chatgpt-visual-ads-dsa-europe
- https://digiday.com/marketing/openai-moves-to-make-chatgpt-ads-more-measurable-and-more-visual/
- https://finance.yahoo.com/media-advertising/articles/openai-putting-visual-ads-next-184547191.html