I think it's trickier than that. People wouldn't trust the answers. At least with ads, you can see that it's an ad. And it seems like there would be liability i

I think it's trickier than that. People wouldn't trust the answers. At least with ads, you can see that it's an ad. And it seems like there would be liability issues to having your AI saying "Use X product because..." vs passively displaying an ad. And since LLMs are non-deterministic, you'd have the risk that it says something that the advertiser wouldn't want. Advertisers would probably be scared of that. I think OpenAI's plugins look like a better fit. Just tell the model what you want and it interacts with services directly. I assume OpenAI is getting a cut of that?