The accuracy of the Mexican National Forest Inventory (NFI) map is derived in four distinct ecogeographical areas, using an assessment design tailored for the project. A main achievement of the design was to integrate the high diversity of classes encompassed at the most detailed subcommunity level of the classification scheme within a cost-controlled statistically sound assessment. A hybrid double sampling strategy was applied to the 2.5 million-ha study area. A total of 5955 reference sites were verified against their NFI map label. The availability of detailed quasi-synchronous reference data for the 2000 Landsat-derived NFI and the high diversity of mapped classes allowed a careful thematic analysis on the selected regions, relevant for national extrapolation. Global accuracy estimates of 64–78 per cent were registered among the four ecogeographical areas (two with mainly temperate climate and the other two with mainly tropical climate), with the lower accuracy levels found in areas more densely covered with forests. According to the estimates, the NFI map tends to underestimate the presence of temperate forest (especially oak) and overestimate the presence of tropical forest in the areas investigated. The analysis of confusions reveals difficulties in unambiguously interpreting or labelling forests with secondary vegetation, herbaceous and/or shrub-like vegetation as well as distinguishing between aquatic vegetation types. The design proved useful from the perspective of accuracy assessments of regional maps in biodiverse regions.