Drylands play a critical role in the global carbon cycle. Recent studies have documented widespread dryland greening largely attributed to CO2 fertilization, yet the role of human activities remained unclear. Here, we used satellite gross primary productivity (GPP) and state-of-the-art dynamic global vegetation models to quantify the global dryland productivity trend from 2001 to 2024 and determine the contributions of climatic and anthropogenic factors. Our results revealed that 29.20% of global drylands experienced significant greening; (P value < 0.1) only 4.91% showed a significant browning trend. The robustness of the satellite GPP trend was corroborated by multiple lines of evidence from leaf area index, enhanced vegetation index, and machine-learning-based upscaling of in situ flux tower GPP (FLUXCOM-X). From 2001 to 2024, the global dryland GPP increased by 1,899 Tg C, with Asia contributing the largest share (983 Tg C). Partial least squares regression analysis revealed that human activities (effect size = 0.68) played a more important role than the CO2 fertilization effect (0.32), with climate factors having only a minor role (0.03). Dynamic global vegetation models from the latest TRENDY v12 project substantially underestimated global dryland greening (multimodel ensemble mean GPP slope = 0.18 +/- 2.45 g C m(-2) y(-1) versus mean satellite GPP slope = 7.66 +/- 4.92 g C m(-2) y(-1)) and overemphasized the role of CO2 fertilization effect and climate. Our results therefore not only advanced our mechanic understanding of global dryland greening but also highlighted the urgent need for refined representation of human land use activities in Earth system models to accurately predict dryland ecosystem dynamics amid global warming and escalating freshwater demand.