Photoacoustic spectroscopy (PAS) has become a popular technique for measuring absorption of light by atmospheric aerosols in both the laboratory and in field campaigns. It has low detection limits, measures suspended aerosols, and is insensitive to scattering. But PAS requires rigorous calibration to be applied quantitatively. Often, a PAS instrument is either filled with a gas of known concentration and absorption cross section, such that the absorption in the cell can be calculated from the product of the two, or the absorption is measured independently with a technique such as cavity ringdown spectroscopy. Then, the PAS signal can be regressed upon the known absorption to determine a calibration slope that reflects the sensitivity constant of the cell and microphone. Ozone has been used for calibrating PAS instruments due to its well-known UV-visible absorption spectrum and the ease with which it can be generated. However, it is known to photodissociate up to approximately 1120 nm via the O 3 + hν (> 1.1 eV) → O 2 ( 3 ∑ g − ) + O( 3 P) pathway, which is likely to lead to inaccuracies in aerosol measurements. Two recent studies have investigated the use of O 3 for PAS calibration but have reached seemingly contradictory conclusions with one finding that it results in a sensitivity that is a factor of two low and the other concluding that it is accurate. The present work is meant to add to this discussion by exploring the extent to which O 3 photodissociates in the PAS cell and the role that the identity of the bath gas plays in determining the PAS sensitivity. We find a 5 % loss in PAS signal attributable to photodissociation at 532 nm in N 2 but no loss in a 5 % mixture of O 2 in N 2 . Furthermore, we discovered a dramatic increase of more than a factor of two in the PAS sensitivity as we increased the O 2 fraction in the bath gas, which reached an asymptote near 100 % O 2 that nearly matched the sensitivity measured with both NO 2 and nigrosin particles. We interpret this dependence with a kinetic model that suggests the reason for the observed results is a more efficient transfer of energy from excited O 3 to O 2 than to N 2 by a factor of 22–55 depending on excitation wavelength.