Glyphosate represents one quarter of global herbicide sales, with growing interest in both its fate in soils and potential to cause non-target phytotoxicity to plants. However, assessing glyphosate bioavailability to plants from soil residues remains challenging. Here we demonstrate that the diffusive gradient in thin-films technique (DGT) can effectively measure available glyphosate across boundary conditions typical of the soil environment: pH 4–9, P concentrations of 20–300 μg P L−1 and NaHCO3 concentrations of 10–1800 mg L−1. In this study, four soils with different glyphosate sorption properties were dosed with up to 16 mg kg−1 of glyphosate and phytotoxicity to wheat and lupin was measured against the DGT-glyphosate concentrations. An improved dose response curve was obtained for root elongation of wheat and lupin across soil types when DGT-glyphosate was used instead of alkaline-extractable (i.e., total extractable) glyphosate. Total extractable glyphosate concentrations of 2.6 and 5.0 mg glyphosate kg−1 in the sandy Tenosol, equivalent to 2.9 and 6.5 μg L−1 DGT-extractable glyphosate, reduced the root length of lupins (but not wheat) by 32–36% compared with the untreated control. DGT is therefore a promising method for assessing phytotoxic levels of glyphosate across different soils.