The FLUXNET dataset contains eddy covariance measurements from across the globe, and represents an invaluable estimate of the fluxes of energy, water and carbon between the land surface and the atmosphere. While there is an expectation that the broad range of site characteristics in FLUXNET result in a diversity of flux behaviour, there has been little exploration of how predictable site behaviour is across the network. Aside from intrinsic interest in this fundamental question, understanding site predictability would be useful for land surface model (LSM) evaluation in setting a priori expectations of model performance. It would also provide a clear rationale for selecting particular FLUXNET sites for model development, evaluation and benchmarking. Here, 155 datasets with 30 minute temporal resolution from the Tier 1 of FLUXNET2015 were analysed in a first attempt to assess individual site predictability. Predictability was defined using the disparity between the ability to simulate fluxes at a site given specific knowledge of the site, and the ability to simulate fluxes given general land surface specifications. We then examined predictability using performance metrics including RMSE, correlation, and probability density overlap, and defined site uniqueness as the disparity between multiple empirical models trained globally and locally for each site. A number of hypotheses potentially explaining site predictability were then tested, including climatology, data quality and site characteristics. We found very few clear predictors of uniqueness across different sites including little evidence that flux behaviour is well discretised by vegetation types. While this result might relate to our definition of uniqueness, we argue that our approach is sound and provides a useful basis for site selection in LSM evaluation.