OpenAI tests visual AI ads in ChatGPT

OpenAI is preparing a new visual ad format for ChatGPT and will expand the tools businesses use to measure their campaigns. The first tests will begin in the United States in late October 2026, during image generation and with an initial group of advertisers.

According to OpenAI, ChatGPT reaches 1.2 billion people every week. The company wants to use that reach to create a revenue stream that can bring the service to more users without relying solely on subscriptions.

Ads alongside images, not inside them

The new format will show images related to products, services, or experiences. The idea is to help you discover a brand and understand how it fits into a specific situation, such as imagining what a piece of furniture would look like in your home or what a wellness service offers.

Ads will initially appear while ChatGPT generates images. They will be clearly identified and separated from the image you asked it to create. OpenAI also says that advertising will not influence the assistant's responses.

That last point matters: an ad may appear in the ChatGPT environment, but it should not change the response you receive to favor a brand.

More tools to measure whether they work

OpenAI is expanding the connections between ChatGPT Ads and the systems businesses already use to analyze sales and conversions. The new integrations include Hightouch, Tealium, and LiveRamp.

It is also adding partners that specialize in attribution, meaning estimating which purchases or sign-ups can be linked to an ad. The list includes AppsFlyer, Triple Whale, Adjust, Northbeam, Branch, Singular, and Kochava, among others.

The company is also exploring geographic tests with Haus, Measured, and WorkMagic. These tests aim to measure whether a campaign causes additional purchases, not just whether it coincides with sales that might have happened anyway.

The first data shared by external partners points to different results depending on the brand and measurement method:

• DV Rockerbox calculated that WeightWatchers had an attributed cost per acquisition 15.3% lower than its combined paid search benchmark.
• WorkMagic reported a statistically significant increase for wellness brand Dose, with 67% of additional purchases coming from new customers.
• Triple Whale reported that 93% of visitors to Portland Leather from ChatGPT Ads were new to the brand.

These are figures from measurement partners, not a guarantee of results for every advertiser.

Controls to protect users and brands

OpenAI says its ads will not appear in emotionally vulnerable, sensitive, or otherwise unsuitable conversations for advertising. It will use automated review, human oversight, and continuous monitoring to make that determination.

Advertisers that meet the requirements will also be able to use Negative Phrases, a feature for excluding specific topics or expressions based on each brand's policies.

In addition, DoubleVerify and Integral Ad Science will test independent tools to assess the safety of the advertising environment. OpenAI says these evaluations will take place without accessing users' private conversations.

For you, the change means ChatGPT will start to look more like a product discovery platform, but with stated limits: ads must be separated from responses and labeled. What remains to be seen is whether those controls hold as the format expands to more countries, more types of queries, and more advertisers.