There is limited understanding about agricultural land dynamic across global drylands and the impacts of different agricultural land transitions on ecosystem productivity changes. This study attempted to fill this gap by examining the agricultural land changes and net primary productivity responses in drylands. A data-driven assessment of the effects on the ecosystem productivity of individual agricultural land displacement was carried out using three methods: the mean difference method, a newly introduced ridge-regression method, and a proposed method based on actual change excluding climate impact. Increases in productivity were accompanied by agricultural land transitions from natural land covers, expected forests. The agriculture expansion mainly replaced sparse vegetation and grassland, and increased the overall productivity of the dryland ecosystem. Mean difference method and ridge-regression method over overestimated the NPP increase caused by agricultural land expansion. Land use change contributed more than 70% to increase in net primary productivity in new agricultural area. But land use change effects less on total productivity in drylands than climate change. Monitoring agricultural land dynamics and distinguishing productivity changes caused by different agricultural land transitions is helpful to the targeted management and sustainable development of agriculture in drylands.