Data quality is spiraling the bottom of the bowl.
That's my concern too. I've been experimenting with MS Copilot. My initial reaction is that it is just a webcrawler that reflects back to you what it reads in simple search engine results. Except sometimes it doesn't get it right, misunderstanding the context or meaning of what it reads.
That means relying on AI is basically like adding an extra person to the old "telephone operator" game, where you whisper a message from one person to another and laugh as the result is completely garbled from the original. My concern is that as the amount of AI content online grows, we may encounter what I call a "recursive hallucination" where AI misunderstanding other AI content creates content that is more and more divorced from reality. And that's without even considering the deliberate biases being introduced by the puppet masters behind each AI.