I'm only partway through, but this article is relevant and interesting: SoK: The Ghost Trilemma (I found it from here). A couple excerpts that caught my attention so far:

We posit the Ghost Trilemma, that there are three key properties of identity—sentience, location, and uniqueness—that cannot be simultaneously verified in a fully-decentralized setting.

And

Due to its inherently adversarial nature, the problem of identifying bots and bad actors online is better addressed than before through advances in adversarial machine learning.

So, what I'm imaging now is an API that monitors the chain and returns three scores for an account (see Table 1 in the document): Sentience, location, and uniqueness. Then:

1. the curator could use that score to decide on their voting. Each curator could decide which factor(s) are more important to them.
2. front-end operators could use that score to give placement priority to accounts with higher scores, and again, each front end could choose alternate priorities for the different factors.

Of course, there'd have to be a funding/business model. Bottom line, I think trying to keep up with the scammers with manual tools is a lost cause. Like so many things, progress depends on bringing together the right mix of funding, motivation, and skills.