Applying fertilizer nitrogen (N) only when a crop response is predicted may enhance use efficiency and profitability while protecting the environment. The crop response index at harvest (RI-harvest, the ratio of the maximum grain yield and that of the plot in question) indicates the actual crop response to applied fertilizer N, although it is calculated after harvest. The objective of this study was to predict RI-harvest of wheat using normalized difference vegetation index (NDVI) response index (RI-NDVI, defined as the ratio of the NDVI in an N-sufficient plot and that in the field in question) captured at Feekes 6 stage. Field experiments were carried out across seven site-years (2017/18 to 2020/21) on wheat. In the first three seasons, the relationships between RI-harvest and RI-NDVI were established by applying a range of fertilizer N levels (0–320 kg N ha− 1), whereas the fourth season was used for validation. The results indicated that RI-NDVI could explain 79% of the variation in RI-harvest using the linear relationship: RI-harvest = 7.077 × RI-NDVI – 6.4885. This model was satisfactorly validated in the fourth season using an independent data set in which a range of fertilizer N doses was applied before the Feekes 6 growth stage. Validation was also carried out by applying a fertilizer N dose corresponding to the predicted RI-harvest. In comparison to the general recommendation, the application of appropriate prescriptive fertilizer N dose along with a fertilizer N dose based on the predicted RI-harvest resulted in an 11% increase in fertilizer N recovery efficiency. It suggests that estimation of in-season RI-NDVI is a viable method for identifying fields that are likely to respond to additional fertilizer N.