Plant–soil feedback (PSF) can influence plant community structure via changes in the soil microbiome. However, how these feedbacks depend on the soil environment remains poorly understood. We hypothesized that disintegrating a naturally aggregated soil may influence the outcome of PSF by affecting microbial communities. Furthermore, we expected plants to differentially interact with soil structure and the microbial communities due to varying root morphology. We carried out a feedback experiment with nine plant species (five forbs and four grasses) where the “training phase” consisted of aggregated versus disintegrated soil. In the feedback phase, a uniform soil was inoculated in a fully factorial design with soil washings from conspecific- versus heterospecific-trained soil that had been either disintegrated or aggregated. This way, the effects of prior soil structure on plant performance in terms of biomass production and allocation were examined. In the training phase, soil structure did not affect plant biomass. But on disintegrated soil, plants with lower specific root length (SRL) allocated more biomass aboveground. PSF in the feedback phase was negative overall. With training on disintegrated soil, conspecific feedback was positively correlated with SRL and significantly differed between grasses and forbs. Plants with higher SRL were likely able to easily explore the disintegrated soil with smaller pores, while plants with lower SRL invested in belowground biomass for soil exploration and seemed to be more susceptible to fungal pathogens. This suggests that plants with low SRL could be more limited by PSF on disintegrated soils of early successional stages. This study is the first to examine the influence of soil structure on PSF. Our results suggest that soil structure determines the outcome of PSF mediated by SRL. We recommend to further explore the effects of soil structure and propose to include root performance when working with PSF.