Soil-steel composite bridges (SSCBs) use the surrounding soil and the culvert’s flexible corrugated steel thin plates to support vertical loads above the culvert. Existing codes use empirical equations based on field measurements and full-scale tests for small and ordinary span SSCBs to estimate the arching process and how it distributes vertical stresses on the culvert walls. The empirical equations then compute the design straining actions. Recent developments in construction technology and urbanization have led to a significant increase in typical spans of SSCBs. This paper investigates large span SSCBs and the associated induced arching action using three-dimensional finite element analysis (FEA). The study compares the FEA findings of straining actions for large span SSCB case studies to the calculations of valid design codes. The comparison demonstrates that existing codes fail to predict the real arching mechanism and, consequently, the resulting straining actions. In addition, the FEA results illustrate how the arching mechanism varies as the span changes. Arching action is typically positive for relatively small spans but becomes negative as the span increases. Finally, results prove that current codes cannot accurately predict real arching, leading to inappropriate design straining actions. For large span SSCBs, these codes require modifications for the arching factors and the profile aspect ratio factors.