Background and Objectives Prediction of haemolytic disease of the foetus and newborn (HDFN) caused by maternal anti‐A/‐B enables timely therapy, thereby preventing the development of kernicterus spectrum disorder. However, previous efforts to establish accurate prediction methods have been only modestly successful. Materials and Methods In a case–control study, we examined 76 samples from mothers and 76 samples from their newborns; 38 with and 38 without haemolysis. The IgG subclass profile of maternal anti‐A and anti‐B was determined by flow cytometry. Samples from newborns were genetically analysed for the A2 subgroup, secretor and FcγRIIa receptor alleles. Results Surprisingly, we found a correlation between the newborn secretor allele and haemolysis (p = 0.034). No correlation was found for FcγRIIa alleles. The A2 subgroup was found only in newborns without haemolysis. Unexpectedly, different reaction patterns were found for maternal anti‐A and anti‐B; consequently, the results were treated separately. For the prediction of haemolysis in A‐newborns, the maternal IgG1 subclass determination resulted in an accuracy of 83% at birth. For B‐newborns, an accuracy of 91% was achieved by the maternal IgG2 subclass determination. Conclusion We improved the prediction of ABO‐HDFN by characterizing maternal anti‐A and anti‐B by flow cytometry and we presented genetic traits in newborns with correlation to haemolysis. We propose a new understanding of A‐ and B‐substances as immunogens that enhance the maternal immune response and protect the newborn, and we suggest that the development of ABO‐HDFN is different when caused by maternal anti‐A compared to maternal anti‐B.