Aim of this work is to study the capability of constructed wetlands to remediate the herbicide terbuthylazine (TER). For this purpose, eight surface-flow wetlands were constructed. Two types of substrates were used, sandy loam and mixture with zeolite. In addition, two levels of plant density were used, two and six rhizomes of Typha latifolia L (10 rhizomes/m2 and 30 rhizomes/m2, respectively). The randomized complete block design was used for statistical evaluation. The results revealed that the constructed wetlands remediate efficiently the herbicide with simultaneous increase of dealkylated metabolites deethyl-terbuthylazine (DET) and deisopropyl-atrazine (DIA) in surface water. The bigger is the density of marsh plant, the more intensive is the reduction of herbicide (7.3 to 23.4%). Regarding the rate of reduction of herbicide during the 44 days of the experiment, the half-life (t1/2) value was twofold in wetlands with 10 rhizomes/m2 of that of 30 rhizomes/m2, i.e., 61 and 30 days, respectively. Moreover, the herbicide and its metabolites were detected in plant tissues, indicating the capability of plant to uptake the xenobiotic substance from the rhizosphere (phytoextraction). The accumulation was greater in foliage samples than in roots. The average uptake of TER was 3219.5 ng/g (n = 8) in the foliage and 1088.6 ng/g (n = 8) in the root. This study clarifies for the first time the possible metabolic pathway by the detection and quantification of the major hydroxy and dealkylated metabolites of herbicide terbuthylazine in the two major wetland compartments such as water and plant tissues of Typha latifolia L. bioreactor.