The capacity to computationally reconstruct premortem DNA methylation from ancient DNA sequences provides a proxy for ancient gene activity patterns, facilitating the study of changes in gene regulation across human populations at different time scales. Presently, due to statistical constraints, existing methods for reconstructing ancient DNA methylation maps are limited to a handful of high-coverage shotgun samples, which comprise a small minority of available ancient samples. The majority of samples are sequenced using in-situ hybridization capture sequencing which targets a predefined set of genomic positions, or low-coverage shotgun sequencing. Here, we develop methods to reconstruct ancient DNA methylation maps for samples that were not subjected to high-coverage shotgun sequencing. To achieve this, we pool together data from multiple individuals to generate a DNA methylation map representative of a population. Our approach is grounded in the observation that variability in DNA methylation is lower within, compared to between, populations. We show that the resulting DNA methylation maps capture meaningful biological information and allow for the detection of differential methylation between human populations. Comparing European populations to populations of Asian origin revealed eight differentially methylated regions, including one within the promoter of IFITM3, which encodes an antiviral protein. Interestingly, a polymorphism in SNP rs12252 within this gene was associated with susceptibility to COVID-19, with the mutant allele being highly prevalent in East Asian individuals and rare in European ones. We also provide guidelines for conducting statistically valid comparative studies involving the DNA methylation of ancient populations. The ability to reconstruct DNA methylation of ancient populations opens up a new frontier in paleoepigenetic research, allowing for the tracking DNA methylation changes throughout human history, using data from thousands of ancient samples.