Here's the entirety of the explanation for "how it works":

Our systems automatically identify content that seems to have little value, low-added value or is otherwise not particularly helpful to those doing searches.

I can only guess that it's some kind of feedback look with time spent on page from Chrome? Maybe? Or an ML system trained to identify the needlessly repetitive and length keyword-laden blather that Google trained everybody to include before the actual recipe / instructions / fact?