I really, really hoped that ads in AI wouldn’t suck. And now… we’re starting to find out.

In this episode, I break down what we’ve learned about how ads are actually showing up inside ChatGPT, following a recent interview with OpenAI’s advertising lead Asad Awan. I share instances where my earlier (admittedly optimistic) predictions were right, what genuinely surprised me, and my concerns, especially if you care about retail media, first-party data, and the future of shopper marketing.

Listen to learn about the architectural separation between ads and the model, contextual targeting vs. surveillance tracking, the “one good ad is enough” philosophy, and what happens when ad surfaces move upstream of the retailer. Most importantly: will AI ads avoid the retail media doom loop, or are we watching it form in real time?

This episode is sponsored by Mirakl Ads

Timeline

[00:00] – Why my City Hall wedding perfectly explains how ads creep into trusted environments.
[01:01] – The OpenAI interview that reveals how ads work inside ChatGPT and why it matters.
[02:30] – The hard wall between the model and the ad layer (and why it's better than I expected).
[05:00] – “One good ad is enough” vs. the saturation playbook of traditional retail media.
[06:00] – The rise of algorithmic, agentic advertising, and why my Spidey sense is tingling.
[08:35] – The real threat to retail media: AI as a high-intent ad surface upstream of the retailer.
[10:30] – AI agents facilitating purchases behind the scenes, and what that means for first-party data.
[11:45] – The year-three test: will revenue pressure break the wall between ads and the model?

Links & Resources

OpenAI Ads Inside ChatGPT: What It Means for Retail Media, Trust & the Future of AI Advertising

Retail Media Breakfast Club