In the Diagrams 2020 conference, we (Mikkel Willum Johansen and myself) reported on our first successes with machine-learning agents identifying and counting diagrams in mathematical papers. One year later, we have progressed, and in this paper I present and discuss ways of creating evidence on the use of diagrams in mathematical publications. Studying a corpus of mathematical journals from the early 21st century and focusing on the intensity of their reliance on diagrams, I explore different means of measuring the use of diagrams as a precursor to further studying their integration into the mathematical argument.