Climate change is expected to increase the frequency of severe droughts, but it remains unclear whether soil biotic conditioning by plant communities with varying species richness or functional group diversity moderate plant–soil feedback (PSF)—an important ecosystem process driving plant community dynamics—under altered rainfall regimes. We conducted a two-phase PSF experiment to test how plant diversity affects biotic PSF under different rainfall regimes. In Phase 1, we set up mesocosms with 15 plant assemblages composed of two grasses, two forbs and two nitrogen-fixing legumes [one, two, three, or six species from one, two, or three functional group(s)] common to the semi-arid eastern Eurasian Steppe. Mesocosms were subjected to two rainfall amounts (ambient, 50% reduction) crossed with two frequencies (ambient, 50% reduction) for a growing season (~3 months). Conditioned soil from each mesocosm was then used in Phase 2 to inoculate (7% v/v) sterilised mesocosms planted with the same species as in Phase 1 and grown for 8 weeks. Simultaneously, the same plant assemblages were grown in sterilised soil to calculate PSF based on plant biomass measured at the end of Phase 2. Feedback effects differed amongst plant assemblages, but were not significantly altered by reduced rainfall treatments within any plant assemblage. This suggests that the examined interactions between plant and soil microbial communities were resistant to simulated rainfall reductions and that increasing plant diversity did not moderate PSF under altered rainfall regimes. Moreover, increasing plant species richness or functional group diversity did not lessen the magnitude of PSF differences between ambient and reduced rainfall treatments. Collectively, these findings advance our understanding of plant diversity's potential to mitigate climate change effects on PSF, showing that in semi-arid grasslands, higher plant diversity may not moderate PSF responses to altered rainfall regimes and highlighting the importance of considering species-specific traits and interaction stability.