I've never seen these two approaches as necessarily in opposition. Yes, search engines can and probably should use proprietary algorithms to derive semantic meaning, relationships, and credibility, but we can also have a more expressive semantic medium to create that content, and give authors more agency when being evaluated by those algorithms. It does not have to be a binary choice between Google black-box and endless viagra spam.

Today Google does consume all sorts of structured data. With low/no code web publishing tools, incompetence is not really a blocker.