ChatGPT will test product image ads beside generated pictures this month
OpenAI is testing a picture-based ad format inside ChatGPT image generation for US Free and Go users. It has not said whether your image prompt picks the ad.
Ask ChatGPT for a picture later this month in the US and a labeled product image may sit right beside it.
OpenAI said on October 5 that it will test a visual ad format inside ChatGPT image generation, starting later this month in the US with an initial group of advertisers. When a user asks ChatGPT for a picture, an image-based ad can appear next to the result. TechCrunch has the announcement. The details below come from outlets that quoted OpenAI's post.
The ad is a picture now
ChatGPT ads so far have been text units under an answer. This format is image-led. The ad images can show "product inspiration, how a product is used, or experiences associated with a service," according to Newscord's roundup of the announcement. Lowyat reports they can run as full-width images or as thumbnail-sized pictures.
On labeling, Lowyat reports that the ads will have "a clear label and remain separate from the image generated by a user." No coverage we read gives the exact label text for this format. OpenAI also repeated its standing line: "As with our other ad formats, advertising does not influence the answers ChatGPT provides" (Engadget).
Who is in the test
Free and Go users in the US. Lowyat lists Plus, Pro, Business, Enterprise, and Edu as excluded. Go is the $8 a month plan, and it has carried ads since ChatGPT's text-ad test began on February 9 (TechCrunch). The same February rules keep ads away from users under 18.
One wrinkle: Android Headlines describes the test as free tier only, while Lowyat and The Next Web both include Go. If you pay for Plus or above, you are outside it either way.
What picks the ad is not public
This is the open question. None of the reports we read, TechCrunch's included, says whether the image prompt itself decides which product appears. OpenAI's post, as quoted, describes the format and the guardrails, not the matching signal.
Two facts are on the record. The placement guardrails "assess whether a conversation is emotionally vulnerable, sensitive or otherwise unsuitable before an ad appears" (The Next Web), so the conversation is an input to the ad system at least as a suitability check. And the text ads that started in February are matched on the conversation's subject, previous chat history, and past ad interactions, with advertisers receiving aggregate views and clicks. Whether image ads use those same signals is the detail to watch. Our July piece on ChatGPT's Ads Manager covers the text-ad side.
Measurement partners and the EU repository
The same announcement added attribution partners including AppsFlyer, Adjust, Branch, Kochava, and Triple Whale, data integrations with Hightouch, Tealium, and LiveRamp, and brand-suitability pilots with DoubleVerify and Integral Ad Science. OpenAI says those independent partners will not get access to private conversations.
The image test is US only, but The Next Web points out that ChatGPT ads already run in 31 European countries and that ChatGPT's August 31 designation as a very large online search engine means a public ad repository by January. That repository must hold "a year of ads, with the advertiser, the runtime, the targeting parameters and the reach." If image ads reach Europe, that is where their targeting parameters would become public.
Why a build studio cares
In a Deep Audit we draw a data-flow map: every system a piece of company or customer data reaches after someone types it. Picture a teammate on a personal Go account pasting a client's unreleased product brief into ChatGPT to mock up packaging. Free and Go conversations have fed text-ad matching since February, so that arrow is already on the map. What is new is the image prompt as an ad surface, and OpenAI has not said whether image prompts feed ad matching. On a personal account, the company has no admin control over either.
The fix is mostly a tier check. Plus, Pro, Business, Enterprise, and Edu sit outside this test, so the question we add to the map is simple: which ChatGPT plan is each person who touches client data actually using?
Next step: read TechCrunch's report, OpenAI's announcement, and The Next Web on the EU ad repository. If your team uses consumer ChatGPT for client work and you want to know where those prompts go, write to us at hello@gattyworks.com.