Thermostable soil protein, known as glomalin, is an important component of soil carbon stocks. Thought to originate from endomycorrhizal fungi, Glomales, this operationally-defined fraction of soil organic matter contains proteins of diverse origin as well as non-protein material, including humic substances. Accumulation results from the balance between production/release and subsequent degradation. Quantification of the protein is subject to uncertainty because of the co-extraction of other components that interfere with the Bradford assay. We studied 10 topsoils from French temperate forests, taken from the national forest monitoring network (Renecofor). Two fractions were extracted, easily extractable (EE) at neutral pH and total extractable (T) at pH 8. Protein was quantified with the colorimetric Bradford method, either by direct calibration using bovine serum albumin (BSA) or by extrapolation of the standard addition plot of BSA. Solubilized organic matter was characterized by using absorbance at 465 and 665 nm and by three-dimensional fluorescence excitation-emission spectroscopy. Neither soil properties nor forest cover influenced glomalin-related soil protein (GRSP) content. Direct assay gave the GRSPEE to be about 1 g kg−1 soil, and GRSPT in the range 3–10 g kg−1, accounting for about 2% of soil organic carbon and about 15% of soil nitrogen. Standard addition plots indicated a two to sixfold under-estimation of protein in total extracts, caused by negative interference with the Bradford assay. The GRSPEE was correlated significantly with both estimates of GRSPT. Under-estimation of GRSPT by direct assay was not related to the E4:E6 ratio but was correlated significantly with the intensity of absorbance at either 460 or 660 nm and with one of the fluorescence peaks. We conclude that GRSPEE is not necessarily more recent than GRSPT and that both fractions may be probes of protein content, but that absolute contents may be under-estimated because of co-extracted humic substances