China released 651 soybean [Glycine max (L.) Merr.] cultivars from 1923 to 1995. However, their diversity is not well characterized. The objective of this study was to quantify genetic diversity in Chinese cultivars via coefficient of parentage (CP), and the relative importance of geographical growing region, province of origin, intended cropping system, era of release, and breeder preferences in determining that diversity. A very low mean CP of 0.02 was found in Chinese soybean cultivars, suggesting the presence of a potentially high level of genetic diversity in Chinese soybean breeding. Cultivar pools from each of the three growing regions of China were almost completely unrelated to each other and exhibited low within‐region mean CP values (<0.06). Similarly, mean CP values within‐ and between‐provinces were low (0–0.2). Cropping systems and release eras also exhibited low within‐ and between‐CP relationships (all <0.07). The low CP values detected here for Chinese soybean breeding resulted from Chinese breeder initiatives to introduce new germplasm into applied Chinese breeding since the 1970s and from a strong tendency to avoid the mating of related parents. Half‐ and full‐sib matings and backcrossing are almost absent from Chinese pedigrees. Although mean CP for cultivars was low, cluster analysis proved to be a surprisingly effective discriminator of diversity patterns. This analysis assigned 270 cultivars to 20 clusters explaining 41% of the total variability in CP. Clusters were almost completely unrelated to each other and could be used as a basis for selection of parents for breeding. Pedigree analysis revealed that more than 30 cultivars grown currently in China trace to U.S. stocks. This successful use of U.S. germplasm in China may provide an important example for future U.S. breeding strategy.