Many gastrointestinal simulation methods have been used to predict bioavailability, but the suitability of different methods for the same metal(loid)s varies widely, which inevitably affects the accuracy of human health risk assessment. Arsenic is a common and important contaminant in many contaminated land situations. It can be readily absorbed and has teratogenic and mutagenic toxicity. Therefore, in this study, four the most commonly used in vitro simulation methods (the Physiologically Based Extraction Test (PBET), In Vitro Gastrointestinal Method (IVG), Soluble Bioavailability Research Consortium (SBRC), the Unified BARGE Method (UBM)) were tested against an in vivo animal live model, to evaluate their effectiveness for the prediction of soil As bioavailability in 10 industrially contaminated soils. The soil As relative bioavailability (RBA) varied between 15% and 68% in the different soils. As bioaccessibility differed between the 4 gastro-intestinal simulation methods. Gastric phase of UBM (UBMG) predicted As relative bioavailability the best of the 4 assays (R 2 = 0.81). This study provides theoretical and technical support to refine human health risk assessment of As in soils from urban industrial legacy contaminated sites.