Almost no δ¹⁸O data are available for leaf carbohydrates, leaving a gap in the understanding of the δ¹⁸O relationship between leaf water and cellulose. We measured δ¹⁸O values of bulk leaf water (δ¹⁸O_LW) and individual leaf carbohydrates (e.g. fructose, glucose and sucrose) in grass and tree species and δ¹⁸O of leaf cellulose in grasses. The grasses were grown under two relative humidity (rH) conditions. Sucrose was generally ¹⁸O-enriched compared with hexoses across all species with an apparent biosynthetic fractionation factor (ε_bio18O_LW, which might be explained by isotopic leaf water and sucrose synthesis gradients. δ18OLW and δ¹⁸O values of carbohydrates and cellulose in grasses were strongly related, indicating that the leaf water signal in carbohydrates was transferred to cellulose (ε_bio = 25.1‰). Interestingly, damping factor p_exp_x, which reflects oxygen isotope exchange with less enriched water during cellulose synthesis, responded to rH conditions if modelled from δ¹⁸O_LW but not if modelled directly from δ¹⁸O of individual carbohydrates. We conclude that δ¹⁸O_LW is not always a good substitute for δ¹⁸O of synthesis water due to isotopic leaf water gradients. Thus, compound-specific δ¹⁸O analyses of individual carbohydrates are helpful to better constrain (post-)photosynthetic isotope fractionation processes in plants.