Fossil pollen data are valuable for understanding past spatio-temporal variations in plant diversity. However, deducing robust patterns of fossil pollen-assemblage properties is challenging due to factors related to pollen production, dispersal, and fossilisation; sediment sampling; and nomenclature and taxonomy of the fossil pollen grains reflecting the vegetation that produced them. If the composition pattern of the most recent fossil pollen assemblage concords with those in modern or surface pollen assemblages from the same location(s), the pattern can be regarded as robust.Using recent (50–500 years) fossil pollen and surface pollen assemblages of angiosperms compiled from the same geographic area of Central Asia, we analyse spatial patterns in compositional turnover and phylogenetic dispersion. These are widely used to quantify different aspects of plant diversity.Despite differences in the data regarding spatial coverage, number of pollen records, and number of pollen taxa, the spatial pattern in the modern and fossil pollen-assemblage properties across climate zones is broadly similar. Especially, the latitudinal patterns of pollen-assemblage properties are strikingly similar between both data types. However, the degree of concordance varies according to the diversity metric being analysed, and the magnitudes of the metric may differ between both data types. When the data are compiled, processed, and standardised following consistent criteria and analysed using appropriate metrics and methods, spatial patterns in pollen-assemblage properties of fossil pollen data are as robust as those in surface pollen data (∼modern vegetation). Therefore, regional patterns of compositional turnover and phylogenetic dispersion can be confidently inferred from older fossil assemblages through the Holocene.