<p>Earthquakes are a major trigger of instability in anti-dip bedding rock slopes (ABRSs). In the evaluation of slope stability under seismic action, the pseudo-static method is still widely used. In this work, mechanical models for flexural toppling and shear sliding failures of ABRSs under seismic loading were established. Corresponding failure mode identification and stability criteria were proposed. A failure surface search strategy based on Adam optimization was then developed. This led to a pseudo-static stability evaluation method (PS-Adam) for ABRSs. The method was validated using two shaking table tests and five numerical models. Additionally, it was contrasted with a pseudo-static approach that uses a genetic algorithm (LEM-GA). The factor of safety and the failure surface under seismic stress are both accurately predicted by PS-Adam, according to the results. The average error in critical instability load relative to shaking table tests is 6.56%. The average error in the factor of safety relative to numerical simulations is 3.42%. Compared with LEM-GA, PS-Adam yields deterministic predictions and significantly higher computational efficiency. The more layers of rock there are, the greater the efficiency gain. For an 800-layer slope, PS-Adam is approximately 220 times faster. For the engineering design of ABRSs and the evaluation of seismic stability, the suggested PS-Adam technique offers a precise and effective tool.</p>