Information on applying different IR techniques to the prediction of nutrients digestibility in feed is limited and no literature is available on the comparison of NIR and MIR spectroscopy for the prediction of protein digestibility in feed or food samples. The objective of this study was to evaluate the potential of using non-invasive vibrational molecular spectroscopy (near IR vs. ATR-FT/MIR) to rapidly determinate intestinal crude protein (CP) digestibility (IPD). The model feed used in this study was wheat. The CP and IPD reference values along with NIR and MIR spectra data were exported to the Unscrambler X v10.4 (CAMO). The raw spectra were preprocessed by 10 different preprocessing algorithms, including baseline offset correction, multiplicative scattering correction (MSC), detrending, first derivative (FD), second derivative (SD), standard normal variate (SNV), first derivative + SNV (FD-SNV), SNV + first derivative (SNV-SD), second derivative + SNV (SD-SNV) and SNV + detrending (SNV-detrending). For CP, the best near-IR model showed an excellent prediction performance (R2 = 0.98); the best ATR-FT/Mid-IR model also gave an excellent prediction performance (R2 = 0.96). Regarding to IPD, the best model obtained by Near-IR technique showed approximate quantitative predictive ability (R2 = 0.68), and the best model generated by ATR-FT/Mid-IR technique obtained similar prediction performance (R2 = 0.67). ATR-FT/Mid-IR models generally showed better predictive abilities than near-IR models, which may be due to the ATR-FT/Mid-IR spectra record fundamental molecular vibrations and can be more easily affected by multiple interferences. The amide I and II bands played important roles in the development of PLS models for CP and IPD. Results from this study demonstrated the potential of using IR spectroscopy for the prediction of nutrient digestibility while more efforts are required to improve the performance of near-IR and ATR-FT/Mid-IR spectroscopy in predicting the IPD.