Abstract: Based on the lack of epistemic of site effects when selecting ground motion in structural seismic vulnerability analysis, a ground motion determination method is proposed with a view towards considering the site effect. The method combines the amplitude modulation of bedrock ground motions and earthquake response analysis of soil layers. The standard input ground motion for seismic vulnerability analysis of structures is established upon this method. Meanwhile, three groups of ground motion input with 'unconsidered', incomplete consideration, and existing methods considering site effects are designed as comparisons. Three kinds of lumped mass simplified models of interlayer force-displacement restoring force relationship were used to analyze structural seismic vulnerability under different ground motion input conditions. Obtained is a comparative analysis for the relative difference ratios of the median m_RR,D and logarithmic standard deviation β_RR,D of the four limit states of the seismic fragility curves of the control group and the standard group considering the three structures. The research shows that the seismic vulnerability of structures with different site effect is different, and that the difference tends to increase with the increase of the limit state of seismic vulnerability. According to the analysis results, the epistemic uncertainty coefficient values considering the site effect in different degrees are developed by the interval estimation model of seismic vulnerability function. The method can provide a reference for considering site effect epistemic uncertainty coefficient correction in seismic vulnerability analysis.