Forests, since their emergence during the Middle Devonian, have profoundly altered global terrestrial ecosystems, climate and environment. A series of qualitative understandings of forests in geologic history have been acquired in light of comprehensive paleobotanical study on in situ fossil forests around the world. However, the community structure and dynamic evolution of fossil forests are awaiting to be revealed. Here, we review and quantitatively assess 38 pre-angiosperm fossil forests ranging from the Devonian to Jurassic periods in terms of distribution, forest composition, canopy height, tree density, biomass, forest stand structure, and spatial pattern by paleobotanical data and multiple analyses. During the evolution of fossil forests, our results show that, (1) canopy height increased twice, first in the Late Devonian-Early Carboniferous and second in the Middle Triassic-Late Jurassic; (2) biomass peaked in the Middle Devonian-Late Carboniferous lycopsid forests then stabilized to the extent that is similar to extant forests after the Middle Triassic; (3) tree density reached the highest in the Devonian-Carboniferous, especially in lycopsid forests, but aligned with that in extant forests after gymnosperms being dominated in the Middle Triassic; and (4) spatial patterns keep predominantly aggregated and similar to extant forests. Four evolutionary stages of fossil forests, are identified as, (I) Initial Forest, from Eifelian to Frasnian, with diverse composition, moderate biomass and tree density, and low canopy height; (II) Geocarbon Forest, from late Famennian to early Permian, with lycopsid dominance, and highest biomass and tree density in geologic history; (III) Dawn Forest, from the middle Permian to Early Triassic, with higher canopy height, and lower tree density and biomass. (IV) Preset Forest, from the Middle Triassic to Late Cretaceous epochs, with gymnosperm dominance and increased canopy height. Our quantitative assessments illuminate the community structure and dynamic evolution of forests in deep time, filling the gap in understanding forest ecosystems between extant and fossil forests.