Insulation surface defects are common defects inside GIS. The ultra-high frequency method is the most widely used to detect GIS insulation defects in operation. At present, due to the unclear understanding of the signal change during the defect degradation, the amplitudes of the UHF signals cannot be correlated with the severity of the defect directly, making it difficult to divide the defect degradation stages reasonably. An experiment is conducted to measure the UHF signals during the degradation of insulation surface defects and the change mechanism of the UHF signals is explored through simulation. Finally, according to the simulation results of UHF signals with different degradation stages, a method for dividing the degradation stage of surface defects based on the normalized mean of work function is proposed. It is found that the surface work function is closely related to the number of discharge pulses and the UHF mean value during the degradation. The UHF mean value normalized by the work function can be used as a criterion for dividing the defect degradation stages into three categories low, medium and high. It is verified that the proposed method can make a good division of the insulation surface defect degradation process through the experiment. The research results lay a theoretical foundation for defect severity assessment based on discharge signals.