The objective was to determine the accuracy of a pregnancy test for predicting nonpregnant cattle based on the evaluation of corpus luteum (CL) blood flow at 20 d (CLBF-d20) after timed artificial insemination (TAI). Crossbred Holstein-Gir dairy heifers (n=209) and lactating cows (n=317) were synchronized for TAI using the following protocol: intravaginal implant (1.0 g of progesterone) and 2mg of estradiol benzoate i.m. on d -10, implant removal and 0.526 mg of sodium cloprostenol i.m. on d -2, 1mg of estradiol benzoate i.m. on d -1, and TAI on d 0. On d 20, animals underwent grayscale ultrasonography (US) to locate the CL and color flow Doppler to evaluate CLBF-d20 using a portable ultrasound equipped with a 7.5-MHz rectal transducer. Based only on a visual, subjective CLBF evaluation, the animals were classified as pregnant or not pregnant. On d 30 to 35, blinded from results of the previous diagnosis, the same operator performed a final pregnancy diagnosis using US to visualize the fetal heartbeat (gold standard; US-d30). A second evaluator also analyzed the CLBF-d20 in the same animals by watching 7-s recorded videos. Blood samples were collected from a subset of 171 females to determine, by RIA, plasma progesterone (P4) concentrations, which indicate CL function. The final pregnancy outcome (US-d30) was retrospectively compared with the CLBF-d20 diagnoses and then classified either as correct or incorrect. The number of true positive, true negative, false positive, and false negative decisions were inserted into a 2 × 2 decision matrix. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the CLBF-d20 test were calculated using specific equations. Binomial variables (pregnancy rate and proportions) were analyzed using Fisher's exact test for the effect of parity and to compare between evaluators and tests (CLBF-d20 vs. plasma P₄). The kappa values were calculated to quantify the agreement between CLBF-d20 and the gold standard (US-d30) and between evaluators. The performance parameters of CLBF-d20 test were as follows: sensitivity=99.0%, specificity=53.7%, positive predictive value=65.1%, negative predictive value=98.5%, and accuracy=74.8%. False negatives represented only 0.4% of the exams. No differences existed in these parameters between evaluators (no. 1 vs. no. 2) and tests (CLBF-d20 vs. plasma P4). Moreover, a high level of agreement was observed between evaluators (0.91). In conclusion, visual evaluation of CLBF-d20 represents a quick, reliable, and consistent diagnostic test that enables the early detection of nonpregnant cattle.