OpenAI has announced a visual ad format for ChatGPT alongside expanded measurement capabilities and brand suitability tools. The announcements arrived around Advertising Week, with some capabilities becoming more widely available while others remain planned tests or early-stage pilots. For advertisers, the story is not just a new place to put an image; it is a developing effort to make ChatGPT advertising easier to evaluate.
Key takeaways
- OpenAI plans to test visual ads during image generation on ChatGPT’s Free and Go plans, starting with a group of US advertisers.
- The ads will carry labels and stay separate from the images users generate.
- Measurement efforts cover conversions, click attribution, brand resonance, and incrementality, but these capabilities are at different stages of development.
- Brand suitability pilots with Integral Ad Science and DoubleVerify are early-stage; the Negative Phrases tool is becoming more widely available.
- Marketers should match each test to a specific business question rather than treat the announcements as proof of performance.
Visual ads enter the image-generation experience
The new format will let advertisers use images to offer product inspiration or demonstrate how a product is used. OpenAI’s initial test will place these ads in the experience of users generating images through ChatGPT’s Free and Go plans.
That context matters for creative planning. A visual that demonstrates a use case has a different job from one that simply announces a brand’s existence. For practitioners considering the format, the useful question is: what can this image communicate that a product name or short description cannot?
OpenAI says the ads will be labeled and distinct from the generated images. That is an important boundary. The announcement describes a separate advertising format, not paid products being inserted into a user’s requested image.
At the time of the announcement, OpenAI planned to begin testing later that month with a cohort of US advertisers. This is a test launch, not evidence of universal availability. Creative teams can prepare concepts, but media teams should confirm eligibility and access before writing the format into a committed plan.
Measurement is expanding beyond the click
OpenAI’s measurement work spans several questions advertisers typically need answered: did someone click, did they convert, did the advertising affect brand perceptions, and did it drive an outcome that would not otherwise have happened?
Those are related questions, not interchangeable ones.
The company’s measurement partners include Kochava, Tealium, and Triple Whale, with offerings designed to help advertisers assess conversions and click attribution. This builds on OpenAI’s own Conversions API and pixel tools, which it introduced for ChatGPT ads in May.
OpenAI is also testing measurement of brand resonance with Kantar and Cint. Separately, it is developing incrementality measurement tools, although work with platforms including Haus and WorkMagic remains in its early stages.
For marketers, the distinction is essential. Click attribution can help connect an interaction with a recorded outcome. Incrementality asks the tougher question: whether the advertising caused additional outcomes. An attributed conversion should not automatically be presented as an incremental one.
The announcement broadens the measurement agenda, but it does not establish that every advertiser can access every capability today. Nor does it provide performance results. The sensible response is to investigate what can actually be measured in a proposed campaign, with which partner, and under what test conditions.
Brand suitability has two separate tracks
OpenAI is developing brand suitability measurement pilots with Integral Ad Science and DoubleVerify. Both efforts are still at an early stage.
Meanwhile, the company is expanding access to Negative Phrases. This tool lets brands prevent ads from appearing in response to queries containing words or terms they consider inappropriate.
These developments should not be bundled into one vague promise of protection. Negative Phrases is a control over where an ad may appear based on query language. The partner pilots concern brand suitability measurement. One does not automatically establish the scope or effectiveness of the other.
For a brand team, the immediate task is to translate suitability policies into decisions that can be applied to a campaign. Which terms should rule out an appearance? Which require a more nuanced assessment? Who approves the list?
Marketers should also ask what the emerging measurement pilots will evaluate before relying on them in internal assurances. A pilot is a useful development, but it is not a finished guarantee with a reassuring logo attached.
A growing ad business still needs campaign-level proof
The visual format and measurement announcements are part of a broader expansion. OpenAI began testing ads in the US in February and has since announced advertising partnerships with platforms including Criteo and Pacvue.
It has also expanded ChatGPT ads into 31 additional markets and hired advertising leadership, including global ads boss David Dugan. The company is opening advertising to more regulated verticals and testing multiple placements in some query responses as it increases ad load.
This activity comes as OpenAI builds its advertising business ahead of a potential IPO. “Potential” matters: the source does not establish a completed or guaranteed offering.
The practical takeaway is that ChatGPT’s advertising environment is evolving across several dimensions at once: formats, geographic reach, advertiser eligibility, placements, and measurement. That makes documenting test conditions particularly valuable. A result should be understood in the context of the format and setup that produced it, rather than treated as a timeless verdict on the whole channel.
What this means for marketers
Nifty Funnel’s view: build a learning plan before building a budget argument. The announcements offer reasons to investigate ChatGPT ads, not reasons to skip evaluation.
Choose one primary question. A conversion-focused test needs a different measurement plan from a brand-resonance study. Decide what the campaign must demonstrate before selecting the tools.
Separate accessible capabilities from promised ones. Ask which formats, integrations, and suitability controls are available to your account. Keep early pilots out of the list of capabilities your campaign depends on unless participation is confirmed.
Brief creative around product understanding. Explore imagery that shows a use case or offers inspiration, consistent with the format’s stated purpose. Do not design around an assumption that the ad becomes part of the user’s generated image.
Set suitability rules before launch. Review Negative Phrases with the people responsible for brand policy, and clarify what any measurement partner will assess.
Define the next decision. Agree on what evidence would justify another test, a budget change, or a pause. Novel inventory can earn attention; continued investment should earn its place through evidence.
FAQ
Where will ChatGPT’s visual ads be tested?
OpenAI plans to test them during image generation on ChatGPT’s Free and Go plans, initially with a cohort of US advertisers. The ads will be labeled and separate from generated images.
What measurement capabilities is OpenAI developing?
Its efforts include conversion measurement, click attribution, brand-resonance testing, and incrementality measurement. Availability varies, and the incrementality work described in the announcement remains early-stage.
How can brands manage unsuitable query contexts?
OpenAI is making Negative Phrases more widely available to prevent ads from appearing against queries containing specified terms. It is also developing early brand suitability measurement pilots with Integral Ad Science and DoubleVerify.
