Background and aims Brachystegia is a species-rich tree genus found in tropical Africa and a typical element of Miombo woodlands, a widely distributed subtype of the Zambezian savanna. Plastid DNA was shown to be largely uninformative to assess species phylogenetic relationships due to widespread chloroplast capture among species. Here, we aim to assess the capacity of nuclear ribosomal DNA (rDNA) to clarify the phylogeny of Brachystegia species while accounting for intra-individual site polymorphisms (2ISPs), which are often present in rDNA and potentially phylogenetically informative. Material and methods Genome skimming sequencing on 47 samples representing 27 of the 29 currently recognized Brachystegia species, allowed us to retrieve complete nuclear ribosomal cistrons encoding for 18S, 5.8S, and 25S rRNA genes (35S rDNA). We reconstructed the Brachystegia phylogeny using Maximum Likelihood methods based on the standard substitution model or integrating 2ISPs (GENOTYPE implementation in RAxML-NG). We additionally tested the effect of partitioning the data (one partition for rDNA genes and one for the ITS1+ITS2). We also conducted network inferences (Neighbor-Net splits graph), as a strict bifurcative approach might not properly model topological uncertainty at shallow phylogenetic depth. Key results 2ISPs-aware and standard phylogenetic reconstructions are largely congruent. We identified several well-supported main clades clarifying the species relationships, including two clades of Miombo woodlands species. Miombo Group A includes species with ovoid to globose axillary dormant buds, while Miombo Group B species have flattened ones. Two morphologically close Brachystegia species (B. kennedyi and B. leonensis) found in Guineo-Congolian rain forests form also a robustly supported clade. 2ISPs coding allowed to identify an additional Guineo-Congolian clade (B. eurycoma and B. nigerica). Ribosomal DNA therefore proves more useful to explore the generic phylogeny than plastid DNA but the species relationships within and among the main clades remain poorly resolved, probably due to recent diversification and/or recurrent hybridization, so that the diversification of Brachystegia remains to be more properly characterised. Conclusion Nuclear and plastid phylogenetic reconstructions of Brachystegia species are discordant. Even if not well-resolved, rDNA phylograms and networks are characterized by taxonomic sorting, while we observe a strictly geographic sorting in the plastid dataset. Most of the species relationships remain to be characterized using additional nuclear markers combined with in-depth morphological investigations.