Summary Current theoretical advances integrating eco‐metabolomics into ecological research provide a novel perspective for predicting interactions between plants and their environment. Yet, whether the plant metabolome varies predictably and consistently with functional traits along environmental clines remains largely unknown. We explored shifts in community‐level responses reflected in community‐weighted means, Rao's quadratic entropy and β‐diversity for functional traits (specific leaf area, leaf area, leaf dry matter content and height) and chemical properties (features, classes and structural/compositional diversity) in eight plant communities distributed along a sub‐Saharan aridity gradient. Additionally, we investigated whether community‐level responses to aridity were consistent with intraspecific shifts in α‐ and β‐diversity estimated for functional traits and the metabolome for Artemisia herba‐alba. Phytochemical similarity increased as climates became more arid at both community and intraspecific levels. Specifically, increasing aridity drove shifts towards underdispersion of chemical properties by decreasing feature, class and structural chemical variation. By contrast, aridity had little impact on functional traits. Our study suggests that the plant metabolome responds more predictably to shifts in aridity than functional traits. Importantly, under future scenarios of climate change, increasingly arid climates may act to alter the metabolome at a higher rate than functional traits within plant communities.