At industrial megasites worldwide, the natural seepage of contaminated groundwater from surface-near layers into rivers and river floodplains may give rise to serious environmental hazards. To prevent adverse effects on the environment, affordable and energy-efficient treatment methods for these sites are required. Large-scale constructed wetlands (CW) may represent a promising approach to protect river catchment areas. A pilot-scale horizontal subsurface flow CW planted with Phragmites australis and an unplanted reference plot were investigated for the removal of monochlorobenzene (MCB), 1,4-dichlorobenzene (1,4-DBC) and 1,2-dichlorobenzene (1,2-DCB) from polluted groundwater in Bitterfeld (Germany). At inflow concentrations of 20 mg L^-1 MCB, 0.19 mg L^-1 1,2-DCB and 0.25 mg L^-1 1,4-DCB, the planted system showed high removals of MCB and 1,4-DCB already after 2 m of the flow path with averages of 48% MCB and 38% 1,4-DCB over all depths (71% MCB and 62% 1,4-DCB in the upper layer of the wetland). The unplanted wetland showed its highest mean removals only after 4 m of the flow path with 35% MCB and 32% 1,4-DCB. 1,2-DCB was not removed in either system. Oxygen limitation prevailed in both systems but concentrations were higher in the planted wetland with redox potentials of 65-90 mV (0-65 mV in the unplanted wetland). The presence and activity of aerobic MCB degraders in the wetland was shown via most probable number (MPN) determination and aerobic microcosm experiments. Since plants have a significant effect on chlorobenzene removal, constructed wetlands provide a high potential for the treatment of chlorobenzene-contaminated groundwater.