Our objective was to retrospectively compare pregnant versus nonpregnant cattle in terms of vascular and morphometric changes in corpora lutea between d 12 and 20 following timed artificial insemination (TAI). Crossbred (Gir × Holstein) lactating dairy cows (n = 136) and heifers (n = 111) were bred after synchronizing ovulations using an estradiol plus progesterone (P4)-based protocol. Corpus luteum (CL) characteristics (area, echotexture, blood flow) were recorded at 48-h intervals from d 12 to 20 following TAI using an ultrasound equipped with color Doppler. Blood samples were collected to determine CL function (plasma P4). Pregnancy diagnosis was performed at d 30. Quantitative assessment of colored pixels within the CL was performed using ImageJ software (National Institutes of Health, Bethesda, MD) and echotexture was quantified using custom software. Continuous variables such as luteal tissue area (LTA), CL blood flow (CLBF), adjusted CLBF (ratio LTA:CLBF), mean pixel value (MPV), pixel heterogeneity (HETER), and plasma P4 were analyzed retrospectively as repeated measures (d 12 to 20) in pregnant versus nonpregnant females using PROC MIXED (SAS Institute Inc., Cary, NC). Main effects were pregnancy status, day of cycle, and their interaction. Further analyses used only data from d 16, because this was the earliest time point of deviation between CLBF of pregnant and nonpregnant animals. We created quartiles for each variable and calculated the risk of pregnancy within quartile. Differences were determined using the chi-squared test. Plasma P4 was significantly higher in prospective pregnant versus nonpregnant cattle on d 18 and 20, whereas LTA differed only on d 20. On d 16, CLBF and adjusted CLBF diverged between pregnant and nonpregnant, followed by a progressive reduction in the latter until d 20. Mean pixel value was not affected by pregnancy status, but HETER was lower on d 20 in pregnant than in nonpregnant cattle. Likelihood of pregnancy increased from quartile (Q)1 (lowest values) to Q4 (highest) of CLBF (Q4 vs. Q1, odds ratio = 32.8, 95% confidence interval: 9.6 to 112.1) and adjusted CLBF [Q4 vs. Q1, odds ratio = 25.4, 95% confidence interval: 8.1 to 80.4), whereas a lower risk of pregnancy was observed only for animals within Q1 of plasma P4 [Q4 vs. Q1, odds ratio = 3.1, 95% confidence interval: 1.3 to 7.2). Day 16 quartiles of LTA, MPV, and HETER did not affect odds of pregnancy. In conclusion, we identified distinct CLBF patterns as early as 16 d after TAI and confirmed that CL function is lost by a reduction in blood flow, which precedes physical regression.