The role of livestock grazing in regulating woody cover and biomass in grass-dominant systems is well recognized. However, the way in which woody plant populations in respond when livestock are removed from grazing in the absence of other disturbances, such as fire, remains unclear.We conducted a 10-year, replicated fencing experiment in a sandy semiarid rangeland in northern China (which has a mean annual rainfall of 365 mm), where fires have been actively suppressed for decades.Fencing dramatically influenced the growth and age structure of the native tree species, Ulmus pumila, which is the sole dominant tree in the area. After a decade, the density of the U. pumila tree population in the fencing plots increased doubly and canopy cover increased triply. The proportion of both saplings (U2) and young trees (U3) increased in fencing plots but decreased in grazing plots after the 10-year treatment period. The effects of fencing on U. pumila trees varied by age class, with potential implications for the future structure of the U. pumila tree community. Decadal fencing led to approximately 80-fold increase in recruitment and a nearly 2.5-fold decrease in the mortality of both U2 and U3. Further, livestock grazing generated a “browsing trap” to the recruitment of both U2 and U3, and had a small impact on the mortality of old trees. A long-term, fencing-driven shift in woody species composition was mediated via its effects on both recruitment and mortality rates.Synthesis and applications. Our results demonstrate that in the long-term absence of both fire and livestock, native woody plant encroachment tends to occur in sandy rangelands, transforming the woody plant demography in the process. The feasibility of full livestock exclusion in sandy rangelands requires further discussion. A balanced amount of livestock grazing may provide critical ecosystem services by regulating woody cover and mediating woody plant encroachment.