The trick. You cannot write “as shown in [3]”, but you need to write “as shown in [3, Theorem 2]” or “as shown in [3, p. 122]”.

Analysis. This does not take too much work. When first citing a result/definition, you need to invest 5 minutes to download the paper and then search for “Theorem” or “Main Theorem”. Advantages of this system:

• Sometimes, the theorem does not say what you wanted, because it has an unfortunate extra assumption. An often repeated comment is that it is hard to parse the cryptic language of old papers, but I think that: (a) old papers are often not so cryptic; (b) if they are, then just can suck it up – you want to avoid a common source of mistakes which is an incorrect identification of two definitions; (c) maybe the different formalism in the old paper is there for a good reason.
• Sometimes, you will find a footnote next to the Main Theorem, which says that “this theorem was already proved before by x”. This can happen even for the most central results from your field, which have been cited countless times, apparently without anybody every opening the paper. It reminds me of the time when I went to my school library to find out what’s so great about Żeromski, but found that the pages were un-cut in his books that were not compulsory reading.
• You will avoid the scourge of “this topic was also studied in [1,5,10,12,13]” and other silly citations. What is “the topic” and how was it “studied”?