Using on-farm tests could improve mastitis treatment decisions. Therefore this study aimed to evaluate on-farm mastitis tests using milk samples submitted for bacteriological culture (BC) with BC as the reference method. Further, the consequences of using on-farm tests to make selective treatment decisions in case of clinical mastitis (CM) were determined. For that 842 milk samples submitted to GD Animal Health laboratory in April–July 2016 were used, each cultured for BC and on the 4 evaluated on-farm tests. Of each test the sensitivity (Se), specificity (Sp), positive predictive value, negative predictive value and accompanying 95%CI’s were determined based on the (dis)agreement with BC for the result categories gram-positive, gram-negative, no growth (NG) and mixed. The Se of the on-farm tests for the result categories gram-positive and gram-negative were high (range 83–92% and 80–89% resp.), as well as the Sp for gram-negative, NG and mixed (range 89–90%, 88–96% and 92–98% resp.). The Sp for gram-positive results was low (range 33–46%) just as the Se for NG and mixed (range 24–46% and 7–18% resp.). Assuming that gram-positive results would be treated with first-choice antibiotics, gram-negative and mixed results with second-choice antibiotics and NG would not be treated with antibiotics, 68–79% of all CM cases would be treated with first-choice antibiotics, versus 47% if the BC result would be used for treatment decisions. The use of second-choice antibiotics would decrease (>6%), but overall more cases would be treated with antibiotics if the decision would be based on the on-farm test result compared with if the BC result would be used. The low Se of the on-farm tests for NG would result in on average 8% of the CM cases left untreated while they had to be treated with an antibiotic. In conclusion, the Se and Sp of the on-farm mastitis tests for gram-negative results were high. The low Sp and Se of gram-positive results and NG resp., could result in an increased number of antibiotic treatments compared with if BC would be available for treatment decisions. However, over 68% of the CM cases would be treated with first-choice antibiotics.