The stratospheric ozone layer plays a key role in atmospheric thermal structure and circulation. Although stratospheric ozone distribution is sensitive to changes in trace gases concentrations and climate, the modifications of stratospheric ozone are not usually considered in climate studies at geological timescales. Here, we evaluate the potential role of stratospheric ozone chemistry in the case of the Eocene hot conditions. Using a chemistry-climate model, we show that the structure of the ozone layer is significantly different under these conditions (4×CO₂ climate and high concentrations of tropospheric N₂O and CH₄). The total column ozone (TCO) remains more or less unchanged in the tropics whereas it is found to be enhanced at mid-and high latitudes. These ozone changes are related to the stratospheric cooling and an acceleration of stratospheric Brewer-Dobson circulation simulated under Eocene climate. As a consequence, the meridional distribution of the TCO appears to be modified , showing particularly pronounced midlatitude maxima and a steeper negative poleward gradient from these maxima. These anomalies are consistent with changes in the seasonal evolution of the polar vortex during winter, especially in the Northern Hemisphere, found to be mainly driven by seasonal changes in planetary wave activity and stratospheric wave-drag. Compared to a preindustrial atmospheric composition, the changes in local ozone concentration reach up to 40 % for zonal annual mean and affect temperature by a few kelvins in the middle stratosphere. As inter-model differences in simulating deep-past temperatures are quite high, the consideration of atmospheric chemistry, which is computationally demanding in Earth system models, may seem superfluous. However, our results suggest that using stratospheric ozone calculated by the model (and hence more physically consistent with Eocene conditions) instead of the commonly specified preindustrial ozone distribution could change the simulated global surface air temperature by as much as 14 %. This error is of the same order as the effect of non-CO₂ boundary conditions (topog-raphy, bathymetry, solar constant and vegetation). Moreover, the results highlight the sensitivity of stratospheric ozone to hot climate conditions. Since the climate sensitivity to strato-spheric ozone feedback largely differs between models, it must be better constrained not only for deep-past conditions but also for future climates.