What are the new ChatGPT ad format and measurement tools for advertisers?
OpenAI confirms a US test of visual ads during image generation, expanded measurement partnerships and brand suitability pilots. Negative Phrases already support qualifying advertisers.
- OpenAI plans a US visual ad test later in October 2026 with an initial advertiser group.
- The first test places labelled ads alongside image generation, separate from the generated image.
- OpenAI confirms expanded conversion, attribution and wider measurement partnerships.
- DoubleVerify and IAS brand suitability evaluation pilots are being developed, not universally available certification.
- Negative Phrases are available to qualifying advertisers for narrow brand specific placement needs.
Personal Take, John IliopoulosEveryone will talk about the visual format. The measurement changes matter more, because they decide whether ChatGPT ads can sit alongside Search and social in a budget conversation or stay a side experiment. OpenAI plans to test labelled visual ads during image generation in the US later in October, with an initial advertiser group. That is a placement test, not a promise that every account can buy it today or that the advert changes the generated image.
Attribution partnerships are useful if they explain what happened, and dangerous if they just give another platform a slice of the same sale. Someone researches in ChatGPT, then converts on a brand search a week later. Both touchpoints may have mattered, but adding the platforms' conversions together does not prove incremental growth. The measurement changes include partner support for conversion data and attribution. I would still check duplicate credit, windows and which integration is actually available for the account.
There are already some brand controls, so targeting, tracking and access are not one completely blank page. OpenAI says Negative Phrases are available to qualifying advertisers for narrow brand specific placement requirements. Its DoubleVerify and IAS evaluation work is still described as pilots, not a finished certification or access to private conversations. I would check eligibility and the actual control offered rather than borrow a menu path from Meta or Google Ads.
A £35 CPL would not excite me if qualified CPL climbed because half those enquiries never got past the first call. On ecommerce, attributed revenue matters less than whether the marginal CPA leaves enough margin. Reaching people mid decision is a genuinely interesting idea. A capped test earns its place when we can tell whether the advertising influenced that decision or simply collected credit once it was already made.
Questions people ask
what is the new ChatGPT visual ad format?
OpenAI plans to test labelled visual ads during image generation, separate from the generated image. The initial test is planned for later in October 2026 in the US with an initial advertiser group, not a universal launch.
how will advertisers measure conversions from ChatGPT ads?
OpenAI confirms expanded partnerships covering conversion data, web and app attribution and wider measurement. Advertisers still need to check the supported integration, attribution window and duplicate conversion handling for their own setup.
what brand suitability controls are available for ChatGPT ads?
Negative Phrases are available to qualifying advertisers for narrow brand specific placement requirements. OpenAI is also developing evaluation pilots with DoubleVerify and IAS, which should not be described as universally available certification or access to private conversations.
Worth a read
Want this level of thinking on your account?
Get a free AI powered PPC audit of your website and goals, with quick wins and a 90 day plan.
Get my free audit
A personal point of view
What I’m seeing, testing and learning in PPC right now.
I’m John Iliopoulos, a PPC and growth marketer with over 10 years in the industry. PPC Supreme is where I share my take on what’s changing, experiments I’m running, things that genuinely work and the lessons behind the results.
