Rain-fed peatlands are dominated by peat mosses (Sphagnum sp.), which for their growth depend on nutrients, water and CO ₂ uptake from the atmosphere. As the isotopic composition of carbon (12,13C) and oxygen (16,18O) of these Sphagnum mosses are affected by environmental conditions, Sphagnum tissue accumulated in peat constitutes a potential long-term archive that can be used for climate reconstruction. However, there is inadequate understanding of how isotope values are influenced by environmental conditions, which restricts their current use as environmental and palaeoenvironmental indicators. Here we tested (i) to what extent C and O isotopic variation in living tissue of Sphagnum is species-specific and associated with local hydrological gradients, climatic gradients (evapotranspiration, temperature, precipitation) and elevation; (ii) whether the C isotopic signature can be a proxy for net primary productivity (NPP) of Sphagnum; and (iii) to what extent Sphagnum tissue δ ¹⁸O tracks the δ ¹⁸O isotope signature of precipitation. In total, we analysed 337 samples from 93 sites across North America and Eurasia using two important peat-forming Sphagnum species (S. magellanicum, S. fuscum) common to the Holarctic realm. There were differences in δ ¹³C values between species. For S. magellanicum δ ¹³C decreased with increasing height above the water table (HWT, R ² =17%) and was positively correlated to productivity (R ² = 7%). Together these two variables explained 46% of the between-site variation in δ ¹³C values. For S. fuscum, productivity was the only significant predictor of δ ¹³C but had low explanatory power (total R ² = 6%). For δ ¹⁸O values, approximately 90% of the variation was found between sites. Globally modelled annual δ ¹⁸O values in precipitation explained 69% of the between-site variation in tissue δ ¹⁸O. S. magellanicum showed lower δ ¹⁸O enrichment than S. fuscum (-0.83 ‰ lower). Elevation and climatic variables were weak predictors of tissue δ ¹⁸O values after controlling for δ ¹⁸O values of the precipitation. To summarize, our study provides evidence for (a) good predictability of tissue δ ¹⁸O values from modelled annual δ ¹⁸O values in precipitation, and (b) the possibility of relating tissue δ ¹³C values to HWT and NPP, but this appears to be species-dependent. These results suggest that isotope composition can be used on a large scale for climatic reconstructions but that such models should be species-specific.