Premise of research: The Cenozoic fossil record is crucial for understanding the evolution of the remarkably high diversity of angiosperms. However, the quality and biases of the angiosperm fossil record remain unclear mainly due to the lack of a global database.

Methodology: We introduce a new global occurrence-based database for Cenozoic angiosperm macrofossils, the Cenozoic Angiosperm Database. We test the temporal, spatial, and phylogenetic biases of the Cenozoic angiosperm macrofossil record and explore their causes.

Pivotal results: The data presented here include 2478 assemblages from all Cenozoic epochs and 1961 sites from all continents, as well as representatives of 221 families (of 445 recognized) and 1859 genera, and show that the Cenozoic angiosperm macrofossil record is extraordinarily rich. However, this rich record is temporally, spatially, and phylogenetically biased: the Miocene is much better sampled than the rest of Cenozoic, the Northern Hemisphere is better sampled than the Southern Hemisphere, and the rosids are better sampled than the rest of the angiosperms. The sampling bias might be caused by collecting effort, geological history, or diverse features of the families, such as growth form and distribution.

Conclusions: The Cenozoic macrofossil record of angiosperms is remarkably rich, especially of woody families found in the Northern Hemisphere. Even if there are numerous biases in these data, a judicious use of the database should be highly informative.