Abstract. The amount of leaves in a plant canopy (measured as leaf area index, LAI) modulates key land–atmosphere interactions, including the exchange of energy, moisture, carbon dioxide (CO2), and other trace gases, and is therefore an essential variable in predicting terrestrial carbon, water, and energy fluxes. The latest generation of Earth system models (ESMs) simulate LAI, as well as provide projections of LAI in the future to improve simulations of biophysical and biogeochemical processes, and for use in climate impact studies. Here we use satellite measurements of LAI to answer the following questions: (1) are the models accurately simulating the mean LAI spatial distribution? (2) Are the models accurately simulating the seasonal cycle in LAI? (3) Are the models correctly simulating the processes driving interannual variability in the current climate? And finally based on this analysis, (4) can we reduce the uncertainty in future projections of LAI by using each model's skill in the current climate? Overall, models are able to capture some of the main characteristics of the LAI mean and seasonal cycle, but all of the models can be improved in one or more regions. Comparison of the modeled and observed interannual variability in the current climate suggested that in high latitudes the models may overpredict increases in LAI based on warming temperature, while in the tropics the models may overpredict the negative impacts of warming temperature on LAI. We expect, however, larger uncertainties in observational estimates of interannual LAI compared to estimates of seasonal or mean LAI. Future projections of LAI by the ESMs are largely optimistic, with only limited regions seeing reductions in LAI. Future projections of LAI in the models are quite different, and are sensitive to climate model projections of precipitation. They also strongly depend on the amount of carbon dioxide fertilization in high latitudes. Based on comparisons between model simulated LAI and observed LAI in the current climate, we can reduce the spread in model future projections, especially in the tropics, by taking into account model skill. In the tropics the models which perform the best in the current climate tend to project a more modest increase in LAI in the future compared to the average of all models. These top performing models also project an increase in the frequency of drought in some regions of the tropics, with droughts being defined as minus one standardized deviation events.