Aerosol radiative forcing represents the largest source of uncertainty for calculating present-day climate sensitivity. This uncertainty has been underestimated in climate models, as they have not adequately accounted for uncertainties in the emissions of aerosols prior to the satellite era. This is especially the case for natural aerosols, such as mineral dust and aerosols from biomass burning. In this paper, we take a step towards reducing this uncertainty by providing a roadmap for improved integration of proxy-based observational constraints on past aerosol variability with model simulations. We first review existing literature on the importance of natural aerosol radiative forcing during historical and paleo-time periods, with a focus on the Last Glacial Maximum and the preindustrial era (ca. 1850) to the present. We then compile existing paleo-data archives and describe their ability to constrain emissions of aerosols, highlighting existing gaps. Model-data combinations are available to estimate radiative effects for dust and volcanoes in several time periods, and for biomass burning for preindustrial to present day, but are not available for other aerosols. Modeled feedbacks of aerosols to climate (W/m²/°C) are 10–100⨉ lower than model-data derived estimates for dust, casting doubt on our ability to successfully model feedbacks in the current generation of Earth system models. Finally, we propose a framework for integrating paleo-archives with models through harmonized datasets, proxy–model translation metrics, and regularized inversion approaches, and identify priority measurements needed to reduce uncertainty in both aerosol loading and radiative-effect efficiency across climate states.