Furfural solvent extraction is a conventional process that removes toxic polycyclic aromatics (PCAs) from eco-friendly processing oil. To increase the product yield and extraction selectivity, the present study introduces water as the co-solvent and investigates its effect using both experimental and theoretical approaches. The effects of the water content in furfural on the product yield, PCA content, aromatic carbon fraction and hydrocarbon type distribution are systematically investigated and optimized. At the appropriate furfural/water ratio in two-stage extraction, the product yield increases by 9 wt% while maintaining the product quality at an acceptable level. An a priori COnductor-like Screening Model for Segment Activity Coefficient (COSMO-SAC) model is used to predict the product composition after extraction. The prediction is consistent with the experimental data. This model can identify the optimal water content using only one experimental result. The result shows that controlling the water content in furfural is critically important during the furfural-extract oil (FEO) furfural extraction process. Both the experimental data and the simulation method in this paper are expected to help control and optimize the extraction process, as well as provide a better understanding of the process. (C) 2018 Elsevier Ltd. All rights reserved.