Abstract: Managing “micropolluted” surface waters is a key challenge for river–lake restoration during the current planning period. Constructed wetlands are widely deployed for such waters because they couple effective pollutant removal with landscape benefits. However, the plant selection and the underlying degradation mechanisms remain unclear. Here, four horizontal subsurface flow constructed wetlands (HSSF-CWs) were established—planted with Iris pseudacorus, Phragmites australis, and Cyperus papyrus, plus an unplanted control—and operated with real micropolluted river water to evaluate plant-specific effects and mechanisms. Results show that the total phosphorus (TP) removal was dominated by substrate adsorption with limited incremental benefit from plants, whereas total nitrogen (TN) and ammonium (NH₄⁺-N) removals were markedly enhanced by vegetation: during the steady phase, planted systems outperformed the control by around 11~26%. Among the species, I. pseudacorus achieved the highest TN removal, while C. papyrus delivered the strongest co-removal of NH₄⁺-N and the permanganate index (COD_Mn). Planting drove the rhizosphere microbiome toward a more specialized, function-enriched state, enriching core degraders such as Sphingomonas and Paracoccus and upregulating pathways related to carbohydrate metabolism, DNA replication/repair, and inorganic ion transport. These findings provide mechanistic evidence to guide vegetation optimization and process intensification of CWs for micropolluted waters.