Some automatic milking systems (AMS) are equipped with sensors to estimate milk components and SCC. However, the accuracy of estimates produced by these sensors is unknown. For these data to be used for herd management, benchmarking and genetic evaluations, AMS estimates should be compared against standardized milk recording laboratory analyses. The aim of this study was to characterize estimates from AMS sensors. Milk samples were collected from all milkings (2.7 ± 1.0 milkings per cow) during a period of 24-h on 10 farms using Lely Astronaut A4 AMS. Samples were analyzed for milk components and SCC. Data from the same period on milk production per individual milking and number of milkings were extracted from the AMS software. Milk composition from laboratory analyses, calculated as 24-h average weighted by milk yield, was compared with the 24-h estimate provided by the AMS. Only records based on 3 or more samples were considered in the comparison (n = 501). Cows were divided according to their DIM as DIM1 (5 to 100), DIM2 (101 to 200) and DIM3 (>201). Statistical analysis was done in R Studio using agricolae package. Concordance Correlation Coefficients (CCC) between AMS sensor estimates and laboratory analysis results were 0.61, 0.59, and 0.69 for fat, protein and lactose percentages, respectively. The CCC for SCC and linear score (LS) were 0.52 and 0.32, respectively. Overall mean differences between AMS and laboratory for fat and protein were small (−0.05 ± 0.5% and −0.001 ± 0.23%, respectively) but mean absolute differences for fat, protein and lactose percentages and LS were (0.38 ± 0.31%, 0.17 ± 0.13%, 0.09 ± 0.06% and 1.32 ± 0.84, respectively). Mean difference for fat percent was larger for DIM1 compared with DIM2 and DIM3 (−0.21 vs. 0.10; P < 0.001), whereas for protein the DIM3 had greater difference (−0.15 vs. 0.05; P < 0.001), than DIM1 and DIM2. No differences were found for SCC; however, LS differences were greater for DIM3 compared with the other groups (−0.42 vs. 0.11; P > 0.05). Results suggest that accuracy of the estimates may differ according to the stage of lactation.