Abstract: Benthic macro-invertebrates are widely used as indicators of aquatic ecosystem health assessment; however, their responses to anthropogenic disturbance may be substantially modified by natural habitat gradients, particularly water depth. Based on field surveys conducted from 2021 to 2023, we compared benthic macro-invertebrate community composition, functional feeding groups, and physicochemical water-quality variables between Taihu Lake, a shallow lake, and Qiandao Lake, a deep reservoir, to determine how water depth influences the interpretation of benthic bioassessment metrics. Qiandao Lake had significantly lower total nitrogen and total phosphorus concentrations and higher water transparency than Taihu Lake. Despite its comparatively good water quality, its benthic assemblage was dominated by hypoxia-tolerant, deposit-feeding oligochaetes, particularly Limnodrilus hoffmeisteri and Branchiura sowerbyi, and was characterized by a lower Shannon-Wiener diversity index and a higher biological pollution index. This resulted in an apparently paradoxical pattern of low anthropogenic disturbance but strong biological indications of pollution. In contrast, Taihu Lake was subject to substantial urban and agricultural inputs, with higher nutrient concentrations and algal biomass and more pronounced eutrophication. Nevertheless, wind- and wave-driven mixing in its shallow waters helped maintain oxic conditions at the sediment-water interface and supported the coexistence of scrapers (gastropods), filter feeders (bivalves), and deposit feeders (oligochaetes), resulting in greater taxonomic diversity and a more balanced functional composition. These findings indicate that water depth shapes benthic assemblages by altering dissolved oxygen conditions and food-resource composition, favoring tolerance-dominated communities in deep-water habitats and functionally diverse communities in shallow-water habitats. Consequently, applying uniform community-based bioassessment thresholds to deep reservoirs and shallow lakes may introduce systematic bias. In deep reservoirs, conventional benthic metrics may have limited capacity to reflect anthropogenic disturbance; their weighting should therefore be reduced, or greater emphasis should be placed on assemblages from inflow zones. Because current aquatic ecosystem assessment standards in China do not explicitly account for water-depth-related habitat differences, naturally low-diversity deep-water communities may be misclassified as heavily polluted, potentially leading to inappropriate management responses. This study provides empirical support for incorporating habitat-specific reference conditions into adaptive aquatic ecosystem health assessment and management.