OpenAI adds a visual ad format to ChatGPT and expands advertiser measurement

ChatGPT will test visual ads during image generation. Advertisers should separate attributed sales from evidence that ads caused extra purchases.

Oct 5, 2026

OpenAI says it will test visual ads during ChatGPT image generation later this month in the US. Labeled ads will stay separate from generated images. OpenAI says advertising leaves answers independent.

Measurement and placement controls

Hightouch, Tealium and LiveRamp integrations help advertisers send conversion data. Attribution partners support reporting and click attribution. OpenAI is also exploring geographic experiments to measure causal impact.

One reported result illustrates the distinction. DV Rockerbox puts WeightWatchers’ attributed acquisition cost 15.3% below its blended paid-search benchmark.

Qualifying advertisers can use Negative Phrases for placement restrictions. DoubleVerify and Integral Ad Science suitability pilots are in development. They will evaluate safeguards in controlled tests without access to private conversations.

Our assessment

Ask which purchases the ad actually caused

An advertiser deciding whether to move search spending into ChatGPT should ask two separate questions. How many purchases did the reporting system credit to ChatGPT? How many additional purchases happened because the campaign ran? A shopper who would have bought anyway can contribute to the first count. The budget decision depends on the second.

Before increasing spending, ask for the attribution window, the comparison group’s results and uncertainty around any measured lift. A lower attributed acquisition cost alone leaves the causal question open. Compare channels using the same purchase definition and measurement method.

For users, the placement of the label matters. Check whether a product appears inside the generated image or in the separate paid placement before treating it as part of the requested result. For brands, ask what scenarios the suitability tests cover and how failures are reported before relying on them to approve a campaign.

Sources

OpenAI: Building advertising for the way people use AI