Plants with crassulacean acid metabolism (CAM) have long been associated with derived anatomy, such as succulence and thick leaves, but quantitative boundaries between C3 and CAM anatomy have yet to be established. Using newly developed computer vision software to measure anatomy, we combined new anatomical measurements with published data across flowering plants. We then used machine learning and phylogenetic comparative methods to investigate relationships between CAM and anatomy. We found significant differences in photosynthetic anatomy between plants with differing CAM phenotypes. Machine learning based classification was moderately to highly accurate in classifying photosynthetic phenotypes. PGLS regression and phylogenetic threshold analyses revealed that increased CAM activity was significantly linked to increased mesophyll cell size and leaf thickness, and decreased intercellular airspace. Significant differences in photosynthetic anatomy exist between non-CAM and CAM plants, as well as between CAM phenotypes, but more data is needed for robust prediction of photosynthetic phenotypes from anatomy alone. Our findings support the hypothesis that morphological innovations are necessary to perform CAM and to transition between CAM phenotypes.