We report a relatively rare study of a national forest inventory in a megadiverse country with the systematic collection of herbarium specimens. The taxonomic identification of 22,659 herbarium collections from 6942 collection sites of Mexico’s national forest inventory resulted in 1464 native tree species (approximately half of Mexico’s estimated total) in 470 genera and 117 plant families. We compared visual tree-species identifications in the field by hired crews, with much more rigorous identification of submitted (mostly sterile) herbarium samples by experienced taxonomists and specialists at the National Herbarium: for 40% of the 22,659 collections, the species identification from the field was confirmed, for 32%, it was corrected at the herbarium, and 27% had been sent without any identification. The most commonly collected plant families were Fagaceae (oak family, 21.7% of all collections), Fabaceae (legumes, 17.7%), and Pinaceae (pine family, 13.3%). The most commonly collected tree species in six major forest-vegetation types were Pinus leiophylla in “coniferous forest”, Quercus magnoliifolia in “highland broadleaf forest”, Liquidambar styraciflua in “mountainous cloud forest”, Bursera simaruba in “lowland evergreen forest”, Lysiloma divaricatum in “lowland dry forest”, and Parkinsonia microphylla in “xerophilous scrub”. We overlapped the six major forest-vegetation types with Mexico’s 15 mainland floristic provinces, as circumscribed by Rzedowski. This resulted in 75 so-called forest-vegetation provinces, of which 35 had at least 20 collection sites. The similarity of species composition among these 35 forest-vegetation provinces was only 17–34% with the Jaccard community index, and 15–42% with the Morisita-Horn community index. The number of physically undetected species was estimated statistically for the 35 forest-vegetation provinces, which indicates that there are forest-vegetation provinces, where the number of species could be up to 8.8-fold higher than those detected in the present work. Finally, we suggest a method to distribute sites optimally among the country in future forest inventories, such as to minimize the average area represented by each site.