Good question! I'm actively developing some automated tools that sift through Steem posts and provide me with candidate lists. Once I have a list of candidate posts, I go through them by hand and do basic quality and plagiarism checks.

Right now it's pretty crude, just doing basic length checks (and wealth checks on authors). I'm playing around with various heuristics (length is a great one - I'm not interested in manually going through 2-line articles) that will effectively pare the list down to something manageable.