When planning the acquisition of new palaeo proxy data for reconstructing past climates, many factors influence the decision of where the proxies are to be collected. One such influence is the likelihood of recording a significant climate change relative to the modern. Another consideration, which is less often considered, is the desire to target regions in which there is significant uncertainty in numerical model predictions of past climate change. New proxy data in these regions enables a more rigorous test of model simulations, and results in better constraints on the various boundary conditions used to force the models. In this paper we present a methodology for targeting new palaeo proxy data, based on model simulations. We use the terrestrial late Miocene (11.6 to 5.3 Ma) as a case study. The late Miocene climate provides insights into future climate behavior, as it is believed to have been considerably warmer than the modern. We carry out a suite of late Miocene atmosphere-only General Circulation Model simulations, which we initially evaluate relative to an example dataset of terrestrial temperature and precipitation. In terms of targeting future data acquisition, we locate those regions in the model which exhibit the largest change in temperature and/or precipitation relative to the pre-industrial. We also locate those regions which display the largest variability between the model simulations, because new data in these regions are most likely to provide a strict test of the reliability of the model results. Among other regions, the Amazon basin, the Chad basin, central Canada, and northern India are identified as potentially useful areas to collect data.