Clay is inherently anisotropic owing to the sedimentation process under gravity and is generally under anisotropic stress state in-situ. The favorably used rotational hardening generally aims to mimick in-situ stress conditions induced by anisotropic consolidation. However, constitutive models adopting such technique do not account for the inherent anisotropy irrespective of consolidation conditions as well as the interplay between anisotropy and loading direction. Furthermore, a non-oblique yield surface should be obtained at the critical state to satisfy thermodynamic principles, irrespective of the rotational hardening rule. This appears to contradict the characteristic of highly anisotropic fabric at the critical state. In this study, an anisotropic bounding surface model that incorporates rotational hardening and fabric evolution is formulated to simulate the behavior of over-consolidated clay, while addressing the aforementioned inconsistency issue. In addition, the model employs an anisotropic elasticity to consider the influence of the initial fabric at a low stress level. The predictive capacity of the model can be demonstrated by simulating several monotonic tests on clay over a wide range of overconsolidation and anisotropic consolidation stress ratios, and cyclic tests under undrained and drained conditions covering varying stress and strain amplitudes.