The stable carbon isotope composition of cellulose (delta(13)Ccell) in fossil wood is valuable for reconstructing past climatic and ecological changes, on seasonal to decadal timescales. However, extracting cellulose from fossil wood is challenging, leading to a lack of delta(13)Ccell data over deep time; moreover, there is a debate about whether the stable carbon isotope composition of whole wood (delta(13)Cwood) can reliably reflect past paleoclimatic or palaeoecological conditions. Here, we present an improved method for extracting cellulose from fossil wood. We initially used conventional methods to extract cellulose from a fossil wood sample a drill core from the Yuncheng Basin, near the Chinese Loess Plateau; however, we were unsuccessful. Subsequently, we successfully extracted cellulose and recovered 94% of the cellulose after modifying the conventional procedure. This involved increasing the reaction time during lignin removal, reducing the concentration of NaOH solution during hemicellulose removal, and employing multiple centrifugation steps for sample separation instead of a single step. We examined the relationship between delta(13)Ccell and delta(13)Cwood values (n = 136), and the results revealed a positive correlation between them (R2 = 0.51, P < 0.001). This indicates that delta(13)Cwood is a dependable proxy for qualitative paleoclimatic reconstruction. However, the apparent enrichment factor epsilon between delta(13)Ccell and delta(13)Cwood values varied between samples, highlighting the need for caution when using records of delta(13)Cwood for quantitative paleoenvironmental reconstruction.