Kinship estimation is widely used in ecological and evolutionary research, particularly in studies of human genealogy and genome-wide associations. In conservation, restoration, agriculture, and forestry, identifying relationships between individuals can be crucial for successful population management and can provide insight into inheritance patterns. Kinship estimation methods are typically designed for large datasets with hundreds of thousands of single-nucleotide polymorphisms. However, studies of non-model species often use much smaller datasets obtained using reduced-representation sequencing. To evaluate the performance of kinship estimation methods under these circumstances, we applied six algorithms to datasets from six non-model Australian flowering plant species (Acacia terminalis, Acacia suaveolens, Banksia serrata, Banksia aemula, Hakea sericea, and Hakea teretifolia), encompassing 3,390 individuals and 369 families. Our results show different performances of kinship methods on reduced-representation sequence data compared with prior evaluations. PC-Relate, RelateAdmix, and Goudet’s beta dosage exhibited limited precision, KING Homo and KING Robust demonstrated high precision with limited sensitivity, while PLINK displayed variable sensitivity and precision. The sensitivity and precision of the methods were affected in various ways by filtering parameters; each method showed its best performance under different thresholds for minor allele frequency and locus missingness. We also present a case study that illustrates a practical application of the methods, demonstrating how estimates of kinship can inform management of seed production areas of the broadleaf hopbush (Dodonaea viscosa). Based on our findings, we offer specific recommendations for utilizing kinship estimation methods in studies of reduced-representation sequence data from non-model species.