Plant DNA from sediments (sedDNA) are increasingly used to reconstruct the past vegetation composition, which contrasts with the few investigations on the relationship between the plant sedDNA signal and modern vegetation. Here, we applied broad-scale terrestrial plant sedDNA metabarcoding on surface sediments from 201 lakes in eastern and northern Asia to discuss the applicability as well as the limitations of using plant sedDNA metabarcoding for palaeovegetation studies. In total, 381 terrestrial plant taxa were determined with the universal plant primers of trnL g and h. Overall, plant sedDNA approach is able to retrieve major vegetation signals. The composition of plant sedDNA reflect well the vegetation types and related climate characteristics, and it also signals which are the dominant taxa in the vegetation. Our results indicate that plant sedDNA metabarcoding could be a reliable proxy of vegetation composition at a sub-continental scale and along large environmental gradients. But certain drawbacks such as limited taxonomic resolution, biases in the relative abundance of taxa, and a generally high variability of samples from similar vegetation types need to be solved before it can be widely applied to reconstruct palaeofloras.