Water content in polyetherimide (PEI) materials is one of the critical factors influencing their quality. Terahertz time-domain spectroscopy (THz-TDS) technique was employed to detect the THz spectra of PEI samples with varying water content. The time-domain spectra were obtained, along with the corresponding variations in the maximum probe current and absorption coefficients. The intrinsic relationships between the maximum time-domain probe current, its variation, absorption coefficients, immersion time, and actual water content were analyzed. Predictive models based on the relationship between the maximum current variation, mean absorption coefficient, and water content were established through linear regression, and experimental validation was conducted. The research demonstrates that both PEI water content prediction models can effectively predict PEI water content, with linear correlation coefficients reaching 0.993 and 0.991, and root mean square errors of 0.0034 and 0.0078, respectively. The findings of this experimental research can provide valuable reference for future studies on novel non-destructive detection methods for water content in polymer materials.