The rapid accumulation of ancient human genomes from various places and time periods, mainly from the past 15,000 years, allows us to probe the past with an unparalleled accuracy and reconstruct trends in human biodiversity. Alongside providing novel insights into the population history, population structure permits correcting for population stratification, a practical concern in gene mapping in association studies. However, it remains unclear which markers best capture ancient population structure as not all markers are equally informative. Moreover, the high missingness rates in ancient, oftentimes haploid, DNA, may distort the population structure and prohibit genomic comparisons. In past studies, ancestry informative markers (AIMs) were harnessed to address such problems, yet whether AIMs finding methods are applicable to aDNA remains unclear. Here, we define ancient AIM (aAIMs) and develop a framework to evaluate established and novel AIMs-finding methods. We show that a novel principal component analysis (PCA)-based method outperforms all methods in capturing ancient population structure and identifying admixed individuals. Our results highlight important features of the genetic structure of ancient Eurasians and the choice of strategies to identify informative markers. This work can inform the design and interpretation of population and medical studies employing ancient DNA.