For comprehensive estimation of the metal treatment efficiency of bioretention systems, information on metal speciation in the stormwater and the effluent is needed. However, so far, most bioretention studies only considered total metal concentrations. Despite their environmental importance, dissolved metals (defined as fractions < 0.45 μm) have only been evaluated in few studies. This study represents the first bioretention study to subdivide the <0.45 μm fraction further by filtration through a 3 kDa ultrafilter (corresponding to appr. 2–3 nm), thus enabling distinction between particulate, colloidal and truly dissolved metals. Higher bioavailability of the truly dissolved fraction has been indicated by previous research, underlining the importance of this study. Since vegetation and salt in stormwater both may be explanatory variables for metal fractionation, these have been added as factors in the utilized full factorial pilot-scale column experiment. While total metal removal was often >95%, detailed fractionation revealed that Cu and (when no salt was added) Zn removal in the <0.45 μm and <3 kDa fractions was significantly lower. Further, mean concentrations of Cu and (in one treatment) Cd in the <0.45 μm effluent fraction did not meet Swedish receiving water quality guidelines. By calculating the particulate, colloidal and truly dissolved fractions, it was shown that bioretention systems affect metal speciation of Cu and Zn. Colloidal and truly dissolved fractions were mostly prevalent in the effluent rather than the influent. Salt affected metal removal mostly negatively. Fractionation was affected by salt mainly in the influent where it increased the concentrations of Cd and Zn in the truly dissolved fraction (no effects on Cu and Pb fractions). In the effluent, Cu and Zn were only slightly affected by salt. Vegetation had mostly no significant effects on metal removal and fractionation. Further integration of detailed metal fractionation into sampling routines in bioretention research is recommended.