Geogenic arsenic (As) contamination is a known issue affecting groundwater quality worldwide. In heterogeneous aquifers, As mobility results from complex physical and geochemical interactions. Extensive monitoring data are required to reliably assess these underlying processes and the natural As heterogeneity. However, effective groundwater characterization is often hindered by limited data availability, high monitoring costs, and resource constraints. This study exploits an alternative source of geochemical information, aggregating data from monitoring wells of sites under remediation, a pervasive network widespread in urbanized areas. We previously demonstrated that, when properly processed to remove anthropogenic influences, these data can provide meaningful insights into groundwater’s pristine composition. We developed a random forest model to predict the probability of As concentrations exceeding the 10 µg/L regulatory threshold. The method was applied to the shallow aquifer of Ferrara province in the Po Valley (northern Italy), a highly anthropized region with known geogenic As issues. Here, local assessments of As natural background levels are often required to distinguish geogenic from anthropogenic source of contamination in remediation procedures, since provided regional-scale assessments lack sufficient resolution. Our model identified areas with high probability of As exceeding 10 µg/L, mostly close to the Po River delta. As mobilization was linked to natural processes driven by the stratigraphic architecture of the area: widespread peat deposits promote redox reactions associated with organic matter degradation, leading to the reduction of Fe/Mn oxides originating from Apennine sediment sources. This study provides a useful tool for groundwater management, improving chemical composition knowledge through an integrated approach, relevant for both local-scale decision-making and large-scale groundwater quality assessments.