Many captive populations and most natural ones lack genealogical information. In this situation, several coancestry estimators have been proposed to infer coancestry from molecular markers. In this study we carried out simulations to test in which situations two analyzed coancestry pairwise estimators produce accurate estimates of the genealogical coancestry. We simulated 10 or 100 codominant neutral markers with six potential alleles, as well as 120 additive, biallelic loci controlling a quantitative trait. Simulations involved census sizes of 50 or 500 individuals in two different population scenarios: (1) populations with no family structure, where contributions of parents to progeny are free and mating is at random; (2) populations with a familiar structure, where 5 or 50 families are obtained through equal contributions of parents to the next generation (10 offspring per family), and parents of each family are chosen at random or depending on their phenotypic value. Our results provide no evidence to suggest a satisfactory behaviour of the analyzed coancestry estimators, although it improves when 100 markers are available. The estimators are likely to be more suitable to infer coancestry than inbreeding, and their accuracy depends on the variance and dispersion of the genealogical coancestry values. Finally, we deduce that molecular coancestry is as effective as the analyzed estimators in order to compare and sort the coancestries.