OpenAI will begin testing a new visual ad format inside ChatGPT's image generation experience later this month, starting with a limited group of U.S. advertisers. The format will use images to show product inspiration, usage, or experiences associated with advertiser products and services.

The test marks the first time ads will appear inside an active ChatGPT creation tool. Until now, ChatGPT ads have appeared below text responses.

How will the ads work?

OpenAI says the visual ads will be clearly labeled and remain separate from the image a user is creating. Advertising also won't influence ChatGPT's responses, according to the announcement.

That separation will matter. A visual ad appearing while someone is generating visual content could feel more native to the experience than a sponsored placement beneath a text response. How much visual distance OpenAI creates between the two will be worth watching once the test begins.

ChatGPT now reaches 1.2 billion people each week. Use cases vary: some users ask questions or research purchases; others create images, work with files, or use other features. Bringing ads into image generation opens inventory to more of those sessions rather than relying only on the chat interface.

OpenAI hasn't shared details on targeting, pricing, creative requirements, or reporting for the new format. The pilot is closed.

What measurement support is OpenAI adding?

OpenAI is broadening the ways advertisers can connect ChatGPT Ads to their existing measurement stack.

Three integrations go live through Hightouch, Tealium, and LiveRamp, which can pipe conversion data from advertisers' systems into ChatGPT Ads. OpenAI also supports attribution partners AppsFlyer, Triple Whale, Adjust, DV Rockerbox, and Northbeam, covering reporting, click attribution, and Conversions API integrations.

Incrementality is the newest frontier. Haus, Measured, and WorkMagic are working with OpenAI on geo-based experiments designed to measure whether ChatGPT Ads caused additional business results. OpenAI describes that work as early stage.

Early partner readouts shared by OpenAI include:

• DV Rockerbox: WeightWatchers' attributed CPA was 15.3% lower than its blended paid search benchmark.
• WorkMagic: Statistically significant incremental purchase lift for Dose.
• Triple Whale: 93% of Portland Leather visitors from ChatGPT Ads were new.

Those results measure different outcomes. OpenAI cautioned that they should not be read as broader ChatGPT Ads benchmarks.

What brand-suitability controls are coming?

Advertisers with placement restrictions will get more control over where their ads appear.

OpenAI has introduced Negative Phrases for qualifying advertisers with placement requirements tied to brand policies. Existing guardrails assess whether conversations are appropriate for advertising, with safeguards intended to keep ads away from sensitive or emotionally vulnerable contexts.

OpenAI is also developing brand-suitability pilots with DoubleVerify and Integral Ad Science. The partners will evaluate how OpenAI's brand-safety standards are applied, without receiving access to private user conversations.

What should advertisers do now?

Most advertisers won't have access to visual ads yet. OpenAI is starting limited in the U.S., with no timeline for a broader rollout. That makes the format a watch item rather than a planning item for most budgets.

If access expands, the question shifts beyond click-through rates or whether new ads outperform existing ChatGPT placements. Advertisers will want evidence that visual ads reach users or influence behavior that other channels weren't already capturing.

OpenAI's measurement updates are where that case gets tested. Expanded attribution support can connect ChatGPT Ads to conversions. The incrementality work could show whether those conversions would have happened anyway.

A new ad format alone doesn't move budget. The case grows stronger if OpenAI demonstrates that visual ads reach different users, drive incremental results, and give advertisers enough placement control to evaluate both.