I wonder if it wouldnt make more sense to just train a Commentator Neural Net with the pages you collect. It then can "remind" you of similar pages you have seen before, directly in the page, with little floating comments like "..skip..." "..contradicts.. http:/link/x/y/z " "..actual paper link.." "..paper is p-hacked shovelware.." "..authors reputation.."

It could be trained to become your personal pre-reading assistant, similar to information accumulators and pre-evaluators in governments agencies or company hqs.