With the development of performance-based earthquake engineering, the ‘risk-probabilistic’ oriented performance evaluation method has gradually gained the attention of researchers, an important part of which is seismic vulnerability analysis. At this stage, there are different kinds of vulnerability methods, and more researches focus on how to combine probability theory with earthquake engineering, but the reasonable comparison for the accuracy and applicability of different methods still requires further research. Based on the nonstationary random mainshock-aftershock sequences, this paper compares three methods commonly used in seismic vulnerability at this stage: linear fitting method, maximum likelihood estimation, and Monte Carlo method. Then, based on a reinforced concrete frame, a case study is carried out, and the applicability of these three methods as well as the influence of random aftershocks are discussed. Generally speaking, the results obtained by the three methods are similar, and the development trends are relatively consistent, which also proves the effectiveness of these three methods to a certain extent. The Monte Carlo method has a long calculation period, the maximum likelihood estimation is more suitable for the performance level of minor damage, and the linear fitting method is more accurate after excluding the scattered points in the collapse state. After considering non-stationary random aftershocks, the obtained structural vulnerability shows an overall left-shifting trend. If the influence of random aftershocks is not considered, the probabilistic risk caused by earthquake sequences will be greatly underestimated.