Remediating heavy-metal pollution in flowing groundwater is challenging because powder adsorbents may agglomerate or wash away, while static tests do not represent dynamic transport. To address these limitations, a PET-Fe-BDC composite geotextile was prepared by in situ synthesis following PAM pretreatment and evaluated for Pb(II) and Cd(II) retention in continuous soil-column seepage experiments. Compared with unamended soil, the composite-geotextile-amended system markedly retarded the downward migration of both ions. In the 0–5 cm retardation layer, cumulative removal efficiencies exceeded 80% up to t_e′, and the breakthrough times of Pb(II) and Cd(II) increased to approximately 6.9 and 6.4 times those in the unamended system, respectively. Thomas, Yan, Clark, and Yoon–Nelson models were used to describe the observed breakthrough behavior. The Yan model provided the most consistent empirical description across the tested curves. The Thomas model closely described the unamended-soil curves and the composite-system Cd(II) curve but showed weaker agreement for composite-system Pb(II), whereas the Clark and Yoon–Nelson models showed lower fitting performance. Because complete saturation was not reached, model-derived parameters were interpreted comparatively. Overall, the integrated PET-Fe-BDC composite-geotextile-amended soil system substantially delayed Pb(II) and Cd(II) breakthrough under continuous seepage, demonstrating its potential for groundwater-remediation barriers.