You're thinking of this in the classical ad placement sense. If there's an established sense of trust with an AI, advertising can become much more subtle and even more manipulative than it currently is. Ad campaigns can last years, and span many conversations, where brands can be woven into the subconscious much more deeply than a traditional ad could do.
They could then be surfaced whenever the user is most vulnerable to be swayed into making a purchase.
> In your classic scenario, you navigate the SERP and hit various websites, that's a lot of surface area for ads.
Most of which never get seen or clicked on. It's a lot of wasted real estate. Ad conversion can become much more efficient if the system deeply knows the person they're interacting with.
You can ask the same question about social media or any Big Tech product today. The answer is somewhere between convenience, a better product, and inertia ("everyone else uses it", etc.).
If Google comes up with an AI that integrates with Gmail, Docs, Drive, Search, and all their other products, and leverages the amount of information they already know about you, then they can provide a much better UX than any competitor could.
I don't doubt there will be OSS competitors that don't do any mind manipulation to get users to purchase something, but most users won't use them. Just like we have sane alternatives to all Big Tech products today, and they're mostly only used by a niche group of technical users.
In your classic scenario, you navigate the SERP and hit various websites, that's a lot of surface area for ads. In this supposed new situation, you don't browse and scan, you just get the answer. The end. It's a way shorter interaction. I'd imagine you see 1 ad. Instead of dozens now.
Also, if you're never hitting any individual websites anymore, surely they'll become hostile, protect content, put up walls, or push for legislation.
This is not a done deal.