The large soil—atmosphere CO₂ flux contributes to observed interannual variations in atmospheric CO₂ concentrations which are poorly understood. Stable isotopes of CO₂ provide a powerful tool to trace biosphere—atmosphere CO₂ fluxes and gain insights to underlying processes. The application of ¹⁸O in CO₂ to carbon cycle studies still involves uncertainties, and controversy, regarding the ¹⁸O signal of soil-respiration and its coupled a-biotic atmospheric CO₂ invasion effect (the diffusion of ambient CO₂ into soil, followed by partial isotopic equilibration with soil water and retrodiffusion). We used experiments and numerical simulations to verify whether the ¹⁸O of the invasion CO₂ retro-diffusion flux from the soil reflects ¹⁸O equilibration with water at the soil surface, or at some depth. This is significant as only surface soil water becomes highly enriched in ¹⁸O due to evaporation. Sterile soil (no CO₂ production) with known water ¹⁸O was used in a flow chamber. Comparing experimental results with predictions based on two commonly used diffusion models (“Penman” and “Moldrup”) indicated that only the Penman model provided good agreement with observations. The depth of full CO₂–water ¹⁸O equilibration was 2-8.5 cm (mean 6 cm), depending on: (1) soil moisture content and its isotopic composition; (2) temperature, dominating rates of isotopic exchange; (3) CO₂ residence time (replacement of the column air above the soil); (4) soil structure (porosity, tortuosity, grain size; the later influencing water surface area exposed to CO₂ exchange). Using field data from a semi-arid forest site, numerical simulations indicated that the ¹⁸O full equilibrium depth varied between 4 and 8 cm (Jan to Nov), being sensitive to temperature and soil moisture. Deepening of the equilibration depth as the soil dries further limits the effects of ¹⁸O evaporative enrichment at the surface on the isotopic composition of the soil–atmosphere CO₂ flux.