Large-scale planted shrublands are increasingly established in arid regions (aridity index (AI) <= 0.2) to combat desertification and enhance carbon sequestration. However, the ecohydrological functioning of these artificial ecosystems-particularly their physiological responses to climate variability compared to natural zonal vegetation-remains poorly understood. This study integrated a plot-scale field precipitation manipulation experiment (PME) on planted Tamarix ramosissima in arid Northwest China (2020-2021) with a global synthesis of evapotranspiration (ET) patterns across 10 large-statured shrubland sites. The PME revealed that while stand-scale ET responded to precipitation changes, plant transpiration (growing-season total: 21.6 mm) remained physiologically unresponsive, with the stand-level response primarily driven by bare soil evaporation. At the global scale, ET in water-stressed planted shrublands exhibited only a weak positive correlation with precipitation and an anomalous negative correlation with potential evapotranspiration (PET). This negative correlation, distinct from energy-limited riparian systems, signals a fundamentally supply-limited regime, further confirmed by a significant positive correlation with AI. Furthermore, despite ET in these shrublands approximating annual precipitation, they maintained a marginal ecosystem water-use efficiency (1.29 g C kg-1 H2O) comparable to non-water-stressed riparian sites (1.18). Integrating these data sets highlights a fundamental hydrological mismatch: the field-observed hydraulic decoupling of plant transpiration from shallow soil moisture alongside a synthesis-derived strong water-carbon coupling. This functional conflict implies that sustaining high biomass incurs a prohibitive water cost, challenging the feasibility of pursuing ecosystem services beyond basic aeolian hazard control. Consequently, our study underscores that afforestation practices must strictly align with regional hydrological constraints to ensure long-term sustainability.