Near infrared reflectance analysis (NIRA) was assessed as a potential technique to measure metabolizable energy (AMEn) in complete poultry feeds. Eighty poultry feeds (for layers, broilers, and turkeys), with known AMEn values (range; 2.120 to 3.553 kcal/g), were used in the present study. Twenty different ingredients were used in the formulation of the diets. A total collection procedure was used in obtaining the AMEn values (a 3-day adaptation period and a 4-day collection period). From the 80 feeds, 49 were used to develop the NIRA calibration for AMEn and the remaining 31 samples were used to test the accuracy of the calibration. The NIRA calibrations were developed for two types of NIRA spectrophotometers, a scanner and a filter. Different mathematical treatments of the spectra were used before selecting the best set of wavelengths to predict AMEn; log 1/R, first derivative, and second derivative. Based on the standard error of the estimate (SEE), all calibrations showed good accuracy at predicting AMEn in poultry feeds. The best results were found using log 1/R for both the scanner (SEE = 0.058 kcal /g) and the filter instrument (SEE = 0.060 kcal/ g). Principal component analysis, a multivariate statistical technique, was also employed in the predictions of AMEn and resulted in SEE of.080 kcal/g. The NIRA technique seems to have a potential as a fast, nondestructive method of AMEn analysis of poultry feed