Abstract Environmental DNA (eDNA) is recognized as a promising sampling tool for biodiversity monitoring. It has also been proposed as a tool for performing population genetic analyses on target species, but early applications make questionable assumptions, such as assuming that the amount of DNA in the sample is directly related to the number of individuals in the environment. In this work, the power of a new analytical framework for detecting genetic differences among populations, which does not make this assumption, is investigated. This is done by using an AMOVA test on only the presence/absence of haplotypes. A sampling strategy with the potential to increase power, including replicate samples and relying on mild assumptions, is also evaluated. A simulation experiment was used to evaluate the ability to detect differences between three populations. The simulation generated mitochondrial haplotypes from three populations with varying diversity and levels of differentiation, and different sampling schemes and AMOVA tests were used. The information of presence/absence of haplotypes, not their relative frequencies, proved useful to perform an AMOVA and the use of replicate samples ensured a better chance of finding statistically significant population differentiation when it existed. The risk of type I errors was also assessed by performing a simulation experiment with no genetic differentiation and that risk was found to depend on the independence between replicates. These results will lay a path for the development of eDNA as a tool for population genetics, as well as serving as a guideline to design sampling plans for future studies and the interpretation of their results.