The objectives of this study were as follows: (a) to assess the reproducibility of polymorphonuclear neutrophil (PMN) cell counts at five predefined endometrial sites (corpus uteri, left horn base, right horn base, left horn tip and right horn tip) and (b) to determine the agreement for the diagnosis of subclinical endometritis (SE) between the different endometrial sites. Forty milking cows between 28 and 34 days post-partum were enrolled for endometrial sampling using cytobrush technique. Intraclass correlation coefficient (ICC) was calculated to evaluate the reproducibility of PMN counts at different sites. The right horn base was found to have the greatest agreement of PMN counts with the other endometrial sites (ICC = 0.66-0.85). Twenty-eight of 40 cows showed no signs of clinical endometritis and were used for evaluation of agreement for the diagnosis of SE, analysed by using Cohen´s kappa (κ) statistics. Agreement for SE diagnosis with PMN cut-off ≥5% was greatest between the right horn base and the right horn tip (κ = 0.84), and with PMN cut-off ≥18% between the right horn base and left horn tip (κ = 1.0), respectively. The results indicate that the right horn base can be regarded as suitable for cytobrush sampling. The probability to detect an animal positive for SE (PMN ≥ 5%) with a single cytobrush sampling was 51.0%; thus, a second sampling is recommended to improve the accuracy.