Abstract Understanding the mechanisms underlying earthquake-induced landslides and assessing seismic responses are crucial for effective mitigation strategies. Earthquakes typically involve a mainshock followed by aftershocks, posing challenges to structures weakened by the mainshock. Highway slope structures, especially those in unsaturated soft-soil slopes, are vulnerable to aftershocks, amplifying the damage caused by the mainshock-aftershock (MSAS) sequence. While existing research primarily focuses on the effects of mainshocks on certain structures, there is a notable gap regarding the damage sustained by unsaturated slope structures under MSAS conditions. Addressing this gap is vital for comprehensive risk assessment and mitigation. To address these challenges, we propose a stochastic model updating approach for seismic reliability analysis. This approach integrates subset simulation with adaptive Bayesian updating and dimensionality reduction using the Karhunen–Lòeve expansion. Shaking table tests on a slope structure with unsaturated red clay soil are conducted to investigate the effects of matrix suction on performance degradation and failure mechanisms. The results reveal spatial variability in soil property parameters, underscoring the need to incorporate this variability into inverse analyses. Traditional deterministic methods or probability-based approaches may overlook such variability. Also, the results indicated our proposed approach enables effective prediction of seismic responses for unsaturated slopes subjected to MSAS sequences. By considering spatial variability and the effects of matrix suction, our method offers a comprehensive framework for seismic reliability analysis of unsaturated slope structures.