For the first time, an assessment of phenotypic variability and genetic polymorphism was performed on endemic plants Scutellaria tuvensis Juz. growing in Tuva (five populations; Russia). Based on morphological traits of individuals, principal component analysis clustered the individuals into three groups depending on characteristics of their habitats: group 1 turned out to be sampled from beach gravel, group 2 from a detrital cone, and group 3 from coarse rock fragments; this finding was confirmed by specific features of the development of the individuals in these habitats. Using inter-simple sequence repeat markers, high genetic polymorphism was identified at the population level: the proportion of polymorphic loci was 95%, expected heterozygosity 0.221, the absolute number of alleles 1.533, and the effective number of alleles 1.376. Population 3 (P 3) was the most genetically homogeneous; P 5 was characterized by the highest genetic diversity. In an unweighted pair group method with arithmetic mean dendrogram, the studied populations formed two major groups: the first cluster included P 4 and P 5, and the second cluster contained plants collected in P 1, P 2, and P 3. An analysis of the population structure using the STRUCTURE software showed the same result, dividing the sample under study into two subpopulations. The genetic differentiation index among populations was 0.232, and gene flow 1.655. According to analysis of molecular variance, intrapopulation differences accounted for 73% of total genetic diversity.