Cadmium, Cu, Ni, Pb, and Zn removal via soil flushing with tannic acid (TA) as a plant biosurfactant was studied. The soil was treated for 30 h in a column reactor at a constant TA concentration and pH (3%, pH 4) and at variable TA flow rates (0.5 mL/min or 1 mL/min). In the soil leachates, pH, electrical conductivity (EC), total dissolved organic carbon, and metal concentrations were monitored. Before and after flushing, soil pH, EC, organic matter content, and cation exchange capacity (CEC) were determined. To analyze the organic matter composition, pyrolysis as well as thermally assisted hydrolysis and methylation coupled with gas chromatography-mass spectrometry were used. Metal fractionation in unflushed and flushed soil was analyzed using a modified sequential extraction method. The data on cumulative metal removal were analyzed using OriginPro 8.0 software (OriginLab Corporation, Northampton, MA, USA) and were fitted to 4-parameter logistic sigmoidal model. It was found that flushing time had a stronger influence on metal removal than flow rate. The overall efficiency of metal removal (expressed as the ratio between flushed metal concentration and total metal concentration in soil) at the higher flow rate decreased in this order: Cd (86%) > Ni (44%) > Cu (29%) ≈ Zn (26%) > Pb (15%). Metals were removed from the exchangeable fraction and redistributed into the reducible fraction. After flushing, the soil had a lower pH, EC, and CEC; a higher organic matter content; the composition of the organic matter had changed (incorporation of TA structures). Our results prove that soil flushing with TA is a promising approach to decrease metal concentration in soil and to facilitate carbon sequestration in soil.