Approximately one-quarter of global forests are distributed in drylands, and their extent is expected to expand further through large-scale afforestation efforts aimed at mitigating climate change. Yet, systematic comparisons of drought risk between natural and planted forests in these water-limited ecosystems remain ambiguous. Leveraging the Loess Plateau-a flagship region of large-scale afforestation since the 1990s-we assessed drought-induced gross primary productivity (GPP) anomalies across natural and planted forests (2001-2020), quantifying sensitivity, adaptation, and integrated vulnerability. Despite comparable drought sensitivity, natural forests exhibited stronger adaptation with slightly more rapid drought response and marginally higher recovery rate, ultimately resulting in lower integrated vulnerability compared to planted forests. Random forest models revealed that climatic aridity (precipitation, aridity index, AI) and soil conditions (soil moisture, texture, nitrogen) collectively governed dryland forests' drought vulnerability. Specifically, drought vulnerability in both natural and planted forests was lower in drier (AI < 0.4, due to reduced sensitivity) and wetter (AI > 0.5, due to greater adaptation) drylands, with natural forests consistently outperforming planted systems by exhibiting greater adaptation under drier conditions and lower sensitivity under wetter ones. The protective role of the soil was reversed due to decreased soil moisture, increased sand content, and nitrogen depletion, but natural forests were unique in their ability to regain weakly mitigated vulnerability under higher sand content. These findings underscore intrinsic constraints in the drought adaptation of planted forests, necessitating rigorous reassessment of afforestation geo-suitability and adaptive management protocols to ensure long-term dryland forest sustainability.