Correctly estimating the effect of elevated CO2 (eCO2) on biomass production is paramount for accurately projecting agricultural productivity, global carbon balances and climate changes. Plant physiology suggests that eCO2 should result in a strongly positive CO2 fertilisation effect (CFE) via positive effects on photosynthesis and water use efficiency. However, the CFE in CO2 experiments is often constrained because of other factors of which rainfall pattern is particularly important. Here, we apply a generally applicable, empirically derived relationship between the CFE and an index of seasonal rainfall balance (SRB), to identify how historical and projected future rainfall patterns modify the CFE using 25 native grassland sites in south-eastern (SE) Australia as a test case. We found that historical and projected rainfall produced SRBs that varied widely from year-to-year resulting in a CFE that was only positive in about 40% of years, with no or even negative biomass responses in the remainder of years; a finding that is in marked contrast to other studies that have not taken account of relationships between rainfall seasonality and plant responses to CO2. The dependence of the CFE on SRB also means that using the CFE from a specific eCO2 experiment can be misleading as the result will be heavily influenced by the SRB during the period of experimentation but this problem can be avoided by using a robust general relationship of the kind used in this study. Generalisations of grassland biomass responses to the rising CO2 concentration are contextual in terms of the variability in precipitation seasonality; as such, this provides a new lens by which to view aboveground responses to the rising CO2 concentration and fosters a novel approach for cross-site comparisons among experiments.