Not sure why it's downvoted, it's been proved many times
People with anecdata about which ads they see, who think of cell phones as inscrutable magic: "No, they definitely are. I can't (won't) think of any other way they'd infer my interests in these topics. You can't see it happening because they are too smart."
I guess the financial incentive might be there for some shady ad network to do it, but that’s just so much risk to take on for such a tiny gain per infected phone..
1) YouTube can read minds and they know what you just thought about
2) with statistical data from literally billions of people it's not so weird to guess what you might be thinking about and show you a video about it.
Again, if they(your phone? Watch? Toaster?) were listening someone would have found some evidence by now, a trace of voice recordings being sent or even something that looks like it might be voice recordings. Yet we have zero. It's not happening, the much easier and much more probable explanation is that we're already tracked through every online service we ever touch, but also we are linked to people we live with or people we interact with, and for a company that possesses exabytes of data it can analyse it's easier to trawl that than try to sneak out voice recordings out of your phone.