Global historical land use scenarios are widely used to simulate the climatic and ecological effects of changes in land cover; however, reliability evaluation of these scenarios for data on China’s forests is missing. By using a historical document-derived Chinese forest dataset (CHFD) for the years 1700-2000, we evaluated the reliability of data on forests in China over three global scenarios—SAGE (Center for Sustainability and the Global Environment), PJ (Pongratz Julia), and KK10 (Kaplan and Krumhardt 2010)—through trend-related, quantitative, and spatial comparisons. The results show the following: (1) Although the area occupied by forests in China in the SAGE, PJ, KK10, and CHFD datasets decreased over the past 300 years, there were large differences between global scenarios and CHFD. The area occupied by forests in China in the SAGE scenario for 1700-1990 was 20%-40% more than that according to CHFD, and that occupied by forests in the KK10 from 1700 to 1850 was 32%-46% greater than that in CHFD. The difference between the PJ and CHFD was lower than 20% for most years. (2) Large differences were detected at the provincial and grid cell scales, where the PJ scenario was closer to CHFD in terms of total forested area. Provinces with large differences in terms of trend and quantity were 84% and 92% of all provinces, respectively. Grid cells with relative differences greater than 70% accounted for 60%-80% of all grids. (3) These global historical land use scenarios do not accurately reveal the spatiotemporal pattern of Chinese forests due to differences in the data sources, methods of reconstruction, and spatial scales.