Summary Isotopically exchanged phosphorus is difficult to determine in soils that strongly sorb P (so that there is little P in solution) and in soils with large concentrations of colloidal P in soil suspensions. A method is proposed in which anion exchange membranes (AEM) are added to the soil suspension after an initial period of isotopic exchange with 32P‐labelled phosphate ions. Isotopically exchanged P, termed EAEM, is calculated from the ratio of labelled phosphate ions to the total phosphate ions on the membrane. The EAEM was compared with the E value measured in an aqueous soil extract (EWater extract) for 14 soils with different degrees of P sorption. The two methods gave similar results in soils with large P concentrations in an aqueous soil extract. However, EWater extract values significantly exceeded the EAEM values by up to 18‐fold when soluble P was near the determination limit (0.008 mg P l−1). In a second experiment, two Ferralsols received further P from inorganic and plant sources and were incubated for 7 days. Treatment effects on labile P were erroneous as detected by the EWater extract but were significant as detected with the AEM method. Third, EAEM values were followed in a Lixisol and a Ferralsol which received labelled phosphate ions with carrier just before the beginning of a 23‐day incubation. The approximate recovery of added inorganic P in the EAEM value suggested that this method adequately samples labile P in P‐sorbing soils. All these results showed that errors in the determination of E values for soils with very small concentrations of P in the soil solution are reduced using the proposed method.