It makes sense if you know your content isn't nice for humans (e.g. full of ads and tracking stuff) but you want it to rank high anyway.

It's an adversarial process. The search engine is, in theory, trying to rank by usefulness to the user, and the site owner is trying to maximize revenue by lying to the search engine. And the user.

Then, maybe the "crawler" should be an actual PC navigating to the browser, taking a screenshot (or live feed) of the page and processing it with AI.

Eh, Google choose to be identifiable as googlebot and to obey robots.txt for other reasons of "good citizenship", because not everybody wants to be crawled.

I'm generally puzzled by Google's reluctance to do manual intervention in these cases. It's not like this is a secret. Just penalize the whole domain for 60 days every time a prominent site lies to the crawler.

If Google's support is any indication, Google doesn't like to involve humans in their processes. There probably isn't enough humans to do this manual intervention you propose.

There are very many sites where the content you see as a non-logged-in user is different from what you see if you have in your possession an all-important user cookie.