You can compare such things to AI generated music playlists based on user submitted musical preferences. Or say, music videos 'suggested' based on what you've already watched. They're not very good. You just end up getting music that most everybody else listens to. Usually limited to specific genre of music or say a supposed time era. Try listening to 90's rock on one and see how nearly every service recommends the exact same stuff. It's the hi-tech version of those "Best of the 70's" or "Oldies" they used to sell on TV. AI makes for a terrible DJ, though I suppose with Clear Channel owning most stations these days, the bar is not set very high.
When it gets down to it, it's just about money. They're not interested in truly catering to your individual tastes. These companies are interested in addictive behavior models and maximizing ad revenue. I would liken it to the approach public education takes in which learning methods are used. One size fits all. Maximum return on investment. Creativity and true problem solving? An afterthought/by product.
I was coding chat bots for fun in the 90's and most recently about 12 years ago. From what I have seen they have not advanced significantly. More processing power, larger databases, some more convoluted code to fake passing a turing test. But the same old stuff on the inside. But wait, now they can search Google and parse results? Literally the same stuff that was possible 15 years ago and what I and some nerds did for entertainment for $0 of funding. All this coverage has more to do with capital investments than anything, which is all it will amount to in the end. Much like the much-lauded return of VR a few years ago and the metaverse in the last two years, it's the latest spam to get you to throw your money into a blackhole. Parlor tricks. But you don't have to take my word. Wait two or three years and see what comes of it all.