Atmospheric CO2 concentration (pCO2) beyond ice core records is typically based on δ11B derived from planktic foraminifera found in equatorial sediment cores. Here, I applied a carbon cycle model over the Plio- Pleistocene to evaluate the assumptions leading to these reconstructed CO2 concentrations. During glacials times simulated atmospheric pCO2 was unequilibrated with pCO2 in the equatorial surface ocean suggesting a bias of up 35 ppm in δ11B-based CO2 concentration. In the Pliocene, surface ocean pH calculated from δ11B in published studies largely differed between equatorial Atlantic and equatorial Pacific. While this difference readily explains most of the resulting pCO2 offsets between studies, it is not supported by models. The values of an under-constrained second variable (dissolved inorganic carbon or total alkalinity) necessary to calculate pCO2 were according to my results partly inconsistent with chemically possible combinations of the marine carbonate system. The model results suggest an existing glacial/interglacial variability in total alkalinity of the order of 100 μmol/kg, which is rarely applied to proxy reconstructions. Simulated atmospheric CO2 is tightly (r2 >0.9) related to equatorial surface ocean pH. This relationship can be used for consistency checks and it is found that reported atmospheric pCO2 of 450–550 ppm at 3–3.5 Ma are likely too high by 50–100 ppm. However, long-term trends in volcanic CO2 outgassing and the strength of the continental weathering fluxes are still unconstrained allowing for a wide range of possible atmospheric CO2 values across the Plio-Pleistocene.