Based on the mechanism of the aerostatic torsional divergence of suspension bridges, two criteria for stiffness degradation are presented and discussed. Under the action of wind loads, the torsional stiffness of the main cable system would degrade when one cable reaches a softening state, and then only the small portion of torsional stiffness contributed by the bridge deck would remain. After degradation, the structure could be easily overturned since the residual stiffness is generally far from enough to resist the negative aerodynamic stiffness. For a suspension bridge immersed in a wind flow, its stiffness would degrade when the upward vertical motion of either main cable is large enough to unload the vast majority or even all of its tension. For convenience, the vertical displacement at the central point of the main cables can be used as a criterion for stiffness degradation. However, lateral deflections of the main cables and horizontal deflections of the bridge towers could postpone the stiffness degradation; moreover, a suspension bridge generally responds stochastically due to the turbulent nature of the oncoming flow, which disenables a reliable critical vertical displacement for assessing the stiffness degradation. In view of this, another criterion based on the cable length is put forward. The numerical results of the dynamic finite element analysis show that the stiffness degradation and the torsional divergence can be well explained with this criterion. Also, a notable discrepancy is found between the two critical wind speeds obtained respectively from the static and the dynamic finite element simulations, showing that buffeting responses of a long-span suspension bridge are non-negligible in the assessment of its aerostatic torsional instability.