I think the people posting about how it doesn't matter if the content comes from an ML model or not as long as it's quality, are missing the point. These chatbots don't generate good quality content. They generate passable content if it's short form and that's about it.
Try making ChatGPT, 3.5-turbo or 4, produce quality content, even via the API with the penalties for repeating terms set to max, and it'll still reiterate content (unless you tell it to produce short form), the text will be stilted and dry (tell it to add more emotion and it starts to generate content that is unnaturally overly emotional). In general it mostly cobbles together a bunch of of high level points about the topic and doesn't generally go into real detail unless you prompt it for specifics on those details, which means you have to know something about the subject matter to make the prompts, then edit the output, then piece the different outputs together so they flow.
At this point you might as well just write most of the content yourself and only use ChatGPT as tool to answer questions which you then rephrase and work into a coherent article.
Again, this is if you want quality content, which is what Google says they want. Otherwise, what you're getting is basically the kind of seo spam content you get from cheap non-native english speakers just with better spelling and grammar. And I wouldn't think that qualifies as quality content in the eyes of search engines.
Stack Overflow banned posting code generated by ChatGPT because of how bad it usually was. So I can imagine even if Google doesn't specifically try and detect ML generated content, it'll still get filtered out due to the quality.
At least until a much better generative text model is produced.