On 5 October OpenAI announced a new visual ad format. It will first be tested while ChatGPT generates an image, later this month in the US, with an initial group of advertisers. The ads will be labeled and separate from the image. The effect numbers it shows are partners' data.
- The new format shows an ad during image generation in ChatGPT. It is labeled and separate from the image. Test later this month in the US.
- OpenAI says advertising does not influence ChatGPT's answers, and says it keeps ads away from emotionally vulnerable and sensitive conversations.
- The effect numbers come from partners: WeightWatchers 15.3% lower cost per acquisition, Dose 67% net-new customers in incremental purchases, Portland Leather 93% new visitors.
You generate an image and while it is being made, you may see an ad. The test starts later this month, in the US.
Images are a good place for an ad. A person looks at the screen, waits for the result, and is in the middle of making something. The advertiser gets attention, which is hard to hold in a text answer.
These are partner numbers. OpenAI quotes them on their behalf. WeightWatchers is one brand against its own benchmark. For Dose, 67% is a share of the increment, not of all sales. 93% new visitors for Portland Leather means new people on the site, and whether they buy, the page does not say. These are examples, picked for the page.
Both trust claims are OpenAI's: advertising does not influence answers, and sensitive conversations stay ad-free. For the second, the page describes the mechanism, a judgment on whether the conversation is appropriate for advertising. The DoubleVerify and IAS pilots are something else: they are for advertisers and check the environment around the ad. A label and separation from the image are promised. How many people will see them in the test, the page does not say.