Background Long-term occupational polycyclic aromatic hydrocarbons (PAHs) exposure is an important risk factor for cognitive impairment. At the same time, it can also cause a decrease of brain-derived neurotrophic factor (BDNF). However, it is not clear whether BDNF plays a key role in the cognitive impairment of workers caused by occupational PAHs exposure. Objective To analyze the correlation between the levels of PAHs in the plasma of coke oven workers and cognitive impairment, and to explore the possible mediating effect of plasma BDNF level on the relationship between PAHs and cognitive impairment. Methods A cross-sectional survey was carried out to select 138 pairs of workers from a coking plant (exposure group) and an energy plant (control group) in a large enterprise in Taiyuan, and the matching variables included age, education level and smoking status. The basic data of the workers were collected by questionnaire. The cognitive function of the workers was assessed by the Montreal Cognitive Assessment (MoCA) scale. Fasting elbow venous blood was collected, the plasma concentrations of 16 PAHs were determined by high performance gas chromatography-mass spectrometry (GC-MS), and the plasma concentration of BDNF was detected by enzyme-linked immunosorbent assay (ELISA). Multiple linear regression model was used to analyze the relationship between 12 PAHs levels in plasma and MoCA scores, and Bootstrap method was used to analyze the mediating effect of BDNF in the relationship between these two indicators. Results The average (\begin{document}$\bar x \pm s $\end{document}) age of workers in the two groups was (48.46±5.04) years, the length of service was (21.45±9.78) years, and 58.7% of the participants reported their education level at secondary vocational school, high school, and above. The median level (25th and 75th percentiles) [M(P₂₅, P₇₅)] of plasma Σ₁₂PAHs concentration of workers in the exposure group [20.937 (9.454, 38.387) μg·L⁻¹] was significantly higher than that of the control group [9.997 (4.952, 23.770)μg·L⁻¹] and the MoCA score (22.82±2.38) of the exposure group was significantly lower than that of the control group (24.60±5.67) (P<0.05). The plasma BDNF showed no significant difference between the exposure group [(29.99±9.80) μg·L⁻¹] and the control group [(31.48±9.09) μg·L⁻¹](P>0.05). Taking plasma PAHs as independent variable and MoCA score as dependent variable, after adjusting selected covariates, the results of multiple linear regression model showed that the MoCA score changed by −0.020 (95%CI: −0.035 - −0.005) for every 1 μg·L⁻¹ increase of plasma Σ₁₂PAHs. Low (<P₃₃), medium ( P₃₃~P₆₆) and high (>P₆₆) dummy variables were set according to the intertertile boundaries (8.31, 23.76) μg·L⁻¹ of the plasma level of Σ₁₂PAH of the workers. The change of MoCA score in the high concentration group versus the low concentration group was −1.167 (95%CI: −1.866 - −0.467). The MoCA score changed by 0.066 (95%CI: 0.012 - 0.119) for each 1 μg·L⁻¹ increase in plasma BDNF level. The results of dichotomous logistic regression analysis showed that the risk of cognitive impairment in the high concentration group was 2.635 (95% CI: 1.085 - 6.398) times higher than that in the low concentration group. The results of mediating effect analysis showed that plasma BDNF level was an intermediate variable between PAHs and cognitive impairment, and 45.50% of PAHs-induced cognitive impairment risk was mediated by reduced BDNF level. Conclusion The level of plasma PAHs in coke oven workers is negatively correlated with MoCA score, and PAHs may mediate cognitive impairment by reducing plasma BDNF level.