Artemisia frigida (subshrub) communities, which are indicators of grassland degradation, are widespread in overgrazed Eurasian steppes. After 4-6 years of enclosure, the community can recover to an Agropyron cristatum-dominated grass community. Understanding the competitive mechanisms between these two key species provides critical insights for the management of semi-arid steppes, where vegetation dynamics are primarily driven by soil moisture. Nevertheless, how soil moisture distribution mediates above- and belowground competition between A. cristatum and A. frigida remains unclear. To address this, we conducted a pot experiment that simulated natural vertical soil moisture heterogeneity with four soil moisture regimes in two soil layers (0-30 cm and 30-60 cm): uniformly dry (D-D), wet upper/dry lower (W-D), dry upper/wet lower (D-W), and uniformly wet (W-W), using both monoculture and mixed planting methods. Key results showed that (1) A. cristatum was more sensitive to soil moisture regimes than A. frigida. Its above- and belowground biomass were significantly higher under moist treatments (W-W, W-D, D-W) than under drought (D-D), whereas the biomass of A. frigida did not differ significantly among water treatments. (2) Compared with monoculture, mixed planting significantly increased the root-shoot ratio of A. frigida but did not affect that of A. cristatum. (3) Competitive ability differed between aboveground and belowground parts: competitive indices (aggressivity and relative competition intensity) revealed that A. cristatum exhibited stronger aboveground competitiveness under moist treatments, while A. frigida dominated aboveground under drought conditions. However, A. frigida consistently exhibited greater belowground competitive ability than A. cristatum across all water treatments in the mixture. These results emphasize that assessments of grass-shrub competition based solely on aboveground indicators may underestimate the competitive advantage of shrubs. Therefore, integrating belowground competitive processes is essential for accurately predicting grass-shrub competition and succession in semi-arid steppes.