Photoacoustic spectroscopy is a sensitive in situ technique for measuring the absorption coefficient for gas and aerosol samples. Photoacoustic spectrometers (PAS) require accurate calibration by comparing the measured photoacoustic response with a known level of absorption for a calibrant. Ozone is a common calibrant of PAS instruments, yet recent work by Bluvshtein et al. (2017) has cast uncertainty on the validity of ozone as a calibrant at a wavelength of 405 nm. Moreover, Fischer and Smith (2018) demonstrate that a low O2 mass fraction in the bath gas can bias the measured PAS calibration coefficient to lower values for wavelengths in the range 532 – 780 nm. In this contribution, we present PAS sensitivity measurements at wavelengths of 405, 514 and 658 nm using ozone-based calibrations with variation in the relative concentrations of O2 and N2 bath gases. We find excellent agreement with the results of Fischer and Smith at the 658 nm wavelength. However, the PAS sensitivity decreases significantly as the bath gas composition tends to pure oxygen for wavelengths of 405 and 514 nm, which cannot be rationalised using arguments presented in previous studies. To address this, we develop a model to describe the variation in PAS sensitivity with both wavelength and bath gas composition that considers Chappuis band photodynamics and recognises that the photoexcitation of O3 leads rapidly to the photodissociation products O(3P) and O2(X, v > 0). We show that the rates of two processes are required to model correctly the PAS sensitivity. The first process involves the formation of vibrationally excited O3(X ̃) through the reaction of the nascent O(3P) with bath gas O2. The second process involves the quenching of vibrational energy from the nascent O2(X, v > 0) to translational modes of the bath gas. Both of these processes proceed at different rates in collisions with N2 or O2 bath gas species. Importantly, we show that the PAS sensitivity is optimised for our PAS instruments when the ozone-based calibration is performed in a bath gas with a similar composition to ambient air and conclude that our methods for measuring aerosol absorption using an ozone-calibrated PAS are accurate and without detectable bias. We emphasise that the dependence of PAS sensitivity on bath gas composition is wavelength dependent and we recommend strongly that researchers characterise the optimal bath gas composition for their particular instrument.