Reconstructions of past climates in both time and space provide important insight into the range and rate of change within the climate system. However, producing a coherent global picture of past climates is difficult because indicators of past environmental changes (proxy data) are unevenly distributed and uncertain. In recent years, paleoclimate data assimilation (paleoDA), which statistically combines model simulations with proxy data, has become an increasingly popular reconstruction method. Here, we describe advances in paleoDA to date, with a focus on the offline ensemble Kalman filter and the insights into climate change that this method affords. PaleoDA has considerable strengths in that it can blend multiple types of information while also propagating uncertainty. Drawbacks of the methodology include an overreliance on the climate model and variance loss. We conclude with an outlook on possible expansions and improvements in paleoDA that can be made in the upcoming years. ▪ Paleoclimate data assimilation blends model and proxy information to enable spatiotemporal reconstructions of past climate change. ▪ This method has advanced our understanding of global temperature change, Earth's climate sensitivity, and past climate dynamics. ▪ Future innovations could improve the method by implementing online paleoclimate data assimilation and smoothers.