We have calibrated the infrared (IR) method for determining CO2 concentrations in apatite with absolute concentrations obtained through nuclear reaction analysis (NRA). IR data were obtained on double-polished apatite wafers using polarized transmission IR spectroscopy. Due to the various sites and orientations of in apatite, the IR spectra are complicated and do not have the same shape in different apatite samples. Hence, simple peak heights are not used to characterize CO2 concentrations in apatite. The total absorbance (Atotal) was derived using the integrated area under the curves in a given polarized spectral region. Then Atotal is calculated as AE//c + 2AE⊥c. The calibration has been carried out for two wavenumber regions, one with high sensitivity and the other applicable to apatite with high CO2 concentrations. The first calibration is for the fundamental asymmetric stretching at wavenumbers of 1600–1300 cm−1, and the CO2 concentration in parts per million can be obtained as (0.0756 ± 0.0036) Atotal/d where d is sample thickness in centimeters. The fundamental stretching bands are strong and hence sensitive for measuring low CO2 concentrations in apatite, down to parts per million level. The second calibration is for the bands at wavenumbers of 2650–2350 cm−1, and the CO2 concentration in parts per million is (9.3 ± 0.6) Atotal/d where d is sample thickness in centimeters. These bands are weak and hence are useful for measuring high CO2 concentrations in apatite without preparation of super-thin wafers. The anisotropy is significant. The difference between AE//c and AE⊥c can reach a factor of 2.73. Hence, for high accuracy, it is best to use polarized IR to determine CO2 concentrations in apatite. For rough estimation, unpolarized IR spectra may be used by estimating Atotal = 3Aunpol, where Aunpol is the integrated absorbance from unpolarized spectra.