Potassium (K) deficiency is one of the main limiting factors in cotton (Gossypium hirsutum L.) production. To study the mechanism of high K-use efficiency of cotton, a pot experiment was conducted. The experiment consisted of two cotton genotypes differing in K-use efficiency (H103 and L122) and two K-application levels (K0: 0 g (kg soil)–1; K1: 0.40 g (kg soil)–1). Root-hair density and length, partitioning of biomass and K in various organs, as well as K-use efficiency of the two cotton genotypes were examined. The results show that there was no significant difference in K uptake between the two genotypes at both treatments, although the genotype H103 (high K-use efficiency) exhibited markedly higher root-hair density than genotype L122 in the K1 treatment. Correlation analysis indicates that neither root-hair density nor root-hair length was correlated with plant K uptake. Furthermore, the boll biomass of genotype H103 was significantly higher than that of genotype L122 in both treatments, and the K accumulation in bolls of genotype H103 was 39%–48% higher than that of genotype L122. On the other hand, the litter index (LI) and the litter K-partitioning index (LKPI) of genotype H103 were 14%–21% and 22%–27% lower than that of genotype L122. Lastly, the K-use efficiency of total plant (KUE-P) of genotype H103 was comparable with that of genotype L122 in both treatments, but the K-use efficiency in boll yield (KUE-B) of genotype H103 was 24% and 41% higher than that of genotype L122 in K0 and K1 treatments. Pearson correlation analysis indicated that KUE-P was positively correlated with BKPI and negatively correlated with LKPI, while KUE-B was positively correlated with BKPI and boll-harvest index (HIB), and negatively correlated with LKPI. It is concluded that there were no pronounced effects of root-hair traits on plant K uptake of the two genotypes. The difference in K-use efficiency was attributed to different patterns of biomass and K partitioning rather than difference in K uptake of the two genotypes.