Nothing but effort. Nobody wants to constantly baby a project just because someone else may change their code at a moment's notice. Why would you want to comb through someone else's html + obfuscated JavaScript to figure out how to grab some dynamically shown data when there was a well documented publicly available API?
Also NewPipe breaks all the time. APIs are generally stable, and can last years if not decades without changing at all. Meanwhile NewPipe parsing breaks every few weeks to months, requiring programmer intervention. Just check the project issue tracker and you'll see it's constantly being fixed to match YouTube changes.
A routine that just returns "yes" will also detect all AI. It would just have an abnormally high false positive rate.