Activity monitors are becoming a widely used tool for estrus detection in dairy cows. The objective of this study is to assess the intensity and accuracy of estrus detection with automated activity monitoring systems. Three commercial dairy farms with an activity monitoring system (Heatime SCR Inc., or Afi PedoPlus, Afikim) that was used for essentially all inseminations between 50 and 80 d in milk (DIM) were enrolled in this observational study. Herds were visited once weekly. Blood samples were collected at wk 5, 7, and 9 postpartum and from a subset of cows on the day of insemination, to measure serum progesterone concentration. Cows were examined at wk 5 for purulent vaginal discharge (PVD). Lameness and body condition were scored at wk 7. Cows were classified as anovular if all 3 blood samples from wk 5, 7, and 9 had progesterone <1 ng/mL. Overall, 6.7% of cows were anovular (range of 3.6% to 8.2% among farms). Of these cows, 39% had a lameness score of 3 or above (1 to 5 scale) and 29% had PVD. Among cows that reached 80 DIM (n = 419), 14% had not been inseminated by 80 DIM; of these, 12% were anovular, 23% were lame, 21% had PVD and 4% had a BCS of < 2.5%. Among 352 blood samples taken on the day of insemination (all services; 212 signaled by the activity system and 140 by Ovsynch), 8% and 3% had progesterone >1 ng/mL; that is, the cows were not in estrus. In this sample, the accuracy of detection of estrus by AAM was high, but a subset of cows would likely require intervention for timely first AI and only a fraction of these were truly not cyclic.