Active optical sensors (AOSs) measure crop reflectance at specific wavelengths and calculate vegetation indices (VIs) that are used to prescribe variable N fertilization. Visual observations of winter wheat (Triticum aestivum L.) plant greenness and density suggest that VI values may be genotype specific. Some sensor systems use correction coefficients to eliminate the effect of genotype on VI values. This study was conducted to assess the effects of winter wheat cultivars and growing conditions on canopy reflectance, as measured by red or amber normalized difference vegetative indices (NDVIs) derived from AOSs. Variations in NDVI values among three wheat cultivars were measured at three growth stages (Zadoks 31, 37, and 65) during 3 yr at three sites in Poland. GreenSeeker Model 505 (Trimble Navigation Limited, Sunnyvale, CA, USA) and Crop Circle ACS-210 (Holland Scientific, Lincoln, NE, USA) sensors were utilized to measure red and amber NDVIs, respectively. Significant (p < 0.05) differences in both forms of NDVI associated with wheat genotypes were observed across years and sites at Zadoks 31, the time when canopy sensing and N fertilization decisions are often made. Lack of a genotype ´ site interaction for both red and amber NDVIs and the presence of a significant genotype ´ year interaction suggested that (i) canopy greenness and density of the same genotype measured at the same growth stage are likely to be stable across different growing conditions, and (ii) NDVI values for a particular genotype tend to vary more across years than across sites. Red and amber NDVI values were significantly different for the same genotype, at the same growth stage, across years. This suggests that a genotype-specific calibration should not be blindly applied from one season to the next, even if at the same growth stage, and this in turn supports the suggestion that in situ reference or virtual strips could mitigate the effect of genotype variability when prescribing variable-rate fertilizer. Because developing temporally variable correction coefficients is not practical, we strongly recommend that an in situ calibration (based on in-field or a virtual reference strip) is utilized to normalize NDVI across genotypes, years, and sites.