ABSTRACT Climate change is having far-reaching consequences on all living beings, altering ecosystems, habitats, and biodiversity worldwide. Species distributions are shifting or decreasing, with alpine plant species being particularly threatened. Natural population monitoring allows the assessment of the impact of human-induced global changes. However, traditional monitoring strategies based on individual counts may produce delayed signals of biodiversity loss. These approaches overlook the fact that genetic diversity is the fundamental basis for evolutionary processes, as it enables populations to adapt to environmental changes, including those caused by climate change. Here, we draw attention to the use of genetic diversity in monitoring schemes to anticipate negative trends in biodiversity and propose two fundamental methodologies: genomics and the use of herbarium specimens. Firstly, in contrast to genetic markers conventionally used to quantify genetic diversity, such as microsatellite markers, genomic approaches provide a vast amount of data that does not require previous knowledge of the studied organism, making them suitable for the study of non-model species. Secondly, herbaria worldwide serve as excellent sources of plant material for comparative studies across time with their precise chronologically recorded collection data. The accuracy of genetic diversity estimates increases with sample size, therefore a large number of vouchers is ideally required. However, the availability of specimens from the same species and populations in public herbaria is limited. Different strategies to quantify genetic diversity are proposed depending on the number of specimens available and their geographic distribution. Finally, we illustrate the potential of this approach in the most restrictive scenario, where only a few individuals are available, and there is no conspecific reference genome. Even in this restrictive scenario, there are signs of genetic depauperation in an alpine species with a narrow distribution, but not in a widely distributed congeneric.