Enhancement of AMPA receptor (AMPAR) function has emerged as a novel strategy for treatment of depression. Nevertheless, studies on AMPAR function in chronic animal models used to predict antidepressant efficacy are surprisingly lacking. We investigated the role of AMPARs in antidepressant action in an unpredictable chronic mild stress (UCMS) model in BALB/c mice. After 3 wk of UCMS, BALB/c mice developed a number of depressive-like behaviours that were successfully prevented by fluoxetine (20 mg/kg) administration. The AMPAR potentiator LY392098 [N-2-(4-(3-thienyl)phenyl)propyl 2-propanesulfonamide] (5 mg/kg), when administered alone, functioned like classic antidepressants by reducing weight loss, fur deterioration and immobility in the tail suspension test. However, LY392098 did not restore sucrose preference and did not reduce anxiety (marble-burying) in stressed mice. In the same protocol, the AMPAR antagonist GYKI (10 mg/kg) reversed most, but not all, of the antidepressant-like actions of fluoxetine. Thus, the antidepressant-like effects of LY392098 were fully predicted by the AMPAR dependence of effects demonstrated for fluoxetine. Our results demonstrate that, in the UCMS paradigm, AMPAR activation exhibits antidepressant-like activity that relates preferentially to specific depressive-like responses and that those specific responses can be defined by their regulation by AMPAR modulation under conditions of stress.