<p>Earthquakes, extreme rainfall, and other conditions can trigger a considerable number of shallow landslides, posing significant safety hazards. Due to the lack of obvious warning signs before sliding, such landslides are not apparent. Traditional remote sensing images and conventional aerial survey data cannot effectively and accurately reflect deformation-related warning signs of sliding. Thus, investigating the subsurface structural characteristics of slopes has become crucial for studying retrogressive shallow soil landslides. This paper takes the Zhongzhai landslide in Niangniangba town, Qinzhou District, Tianshui city, Gansu Province, China, as the research object. Electrical resistivity profiles of the landslide area were obtained by electrical resistivity tomography (ERT), using in situ light dynamic penetration tests and core drilling to confirm the relationship between resistivity and formation lithology to realize the fusion of multisource data. A three-dimensional model of electrical resistivity was constructed to characterize the stratigraphic structure. Combined with unmanned aerial vehicle (UAV) photogrammetry and on-site investigation to obtain terrain features, a three-dimensional geological model of the Zhongzhai landslide was constructed. The development process and genesis mechanism of landslides at the loess–bedrock interface were explored via numerical simulation. The results demonstrate how the stratigraphic structure and water table influence the development of retrogressive shallow soil landslides. This article can provide a reference for the stability evaluation and prediction of retrogressive shallow soil landslides.</p>