Dairy cattle are affected by Johne's disease. It is caused by Mycobacterium avium subspecies paratuberculosis (MAP). Suboptimal diagnostic tests add more to the productivity loss resulting from this disease. Agreement between and within different commercial kits is crucial in the decision-making process of disease surveillance programmes. This study compared two ELISAs, that is, Johne's disease commercial antibody detection kits (A and B), using milk and serum samples from New Zealand dairy cattle. These results were also compared with a subset of faecal PCR results. Five scenarios were considered for the comparison of ELISA tests. The point estimates of kappa coefficients (k) between the serum (0.84–0.94) assays were higher than the milk assays (0.59–0.82). The point estimates of kappa coefficients between serum and milk ELISA outcomes were higher for kit B (k = 0.79–0.86) than for kit A (k = 0.55–0.79). The point estimates of kappa coefficients between the ELISA and faecal PCR outcomes varied between 0.43 and 0.74. ELISA tests had point estimates of sensitivity ranging from 0.67 to 0.88 and specificity from 0.62 to 0.93, relative to the faecal PCR test. Results suggest that serum provides a better choice of sample type when both commercial kits A and B are used for Johne's disease surveillance of dairy cattle in New Zealand. Milk assays can be cost-effective to diagnose MAP-positive animals; kit B can be best suited for New Zealand conditions, provided the repeatability of the results is validated.