The Stoebe clade is a Cape-centred lineage comprising ca. 70 species, currently in eight genera. Despite recent generic taxonomic changes, many generic characters are ambiguous, and generic circumscription is likely to remain unstable in the absence of a robust species-level phylogeny. Here, we present a hypothesis of species relationships in the clade, based on parsimony and Bayesian analyses of DNA sequence and morphological characters, in order to assess generic monophyly. The trees based on morphology alone and plastid sequences alone are poorly resolved, with few well-supported nodes, while those based on nuclear sequences alone are better resolved. There is no evidence of phylogenetic conflict, and concatenation of all three data sources produces a well-resolved tree with many highly-supported nodes. We show that none of the currently-accepted genera are monophyletic, with the exception of Muscosomorphe, which is supported as monospecific. Amphiglossa species are recovered in two separate clades, requiring the resurrection of the genus Pterothrix to house ten of the twelve species. All the characters used to circumscribe the remaining genera, such as the possession of ray florets, are shown to be homoplasious. As a result, these remaining genera (Dicerothamnus, Disparago, Elytropappus, Myrovernix, Seriphium, and Stoebe) are here synonymised with Stoebe. The main morphological synapomorphy for this enlarged concept of Stoebe is the aggregation of the tiny capitula into conflorescences with varying degrees of complexity and congestion. The proposed taxonomy for the Stoebe clade comprises just four genera: Amphiglossa, Muscosomorphe, Pterothrix, and Stoebe. We present the new combinations, keys and generic descriptions, as well as keys to the species.