The net is vast and infinite and funnels the human id.
Autotegenerated content will never be a clean feed until machine learning evolves.
Or to put it another way, teach your kids to talk with you if they see or experience something icky.
Autogenerating content targeted at kids is "systematic abuse for profit" in the same way that Disney holding focus groups is "abuse". The content is clearly entertaining to kids, or else there would be no economic incentive to create it.
The algorithm occasionally picks up some nasty troll content by accident (hence degrees of separation). It is clearly unintentional as it goes against the financial interests of the channels.
This is a long way from "zomg illuminati bot conspiracy subliminal evil content child abuse".