Cognitive impairment is a core feature of schizophrenia and a strong predictor of psychosocial disability. Auditory-based targeted cognitive training (TCT) aims to enhance verbal learning and other domains of cognitive functioning through "bottom-up" tuning of the neural systems underlying early auditory information processing (EAIP). Although TCT has demonstrated efficacy at the group level, individual response to TCT varies considerably, with nearly half of patients showing little-to-no benefit. EEG measures of EAIP, mismatch negativity (MMN) and P3a, are sensitive to the neural systems engaged by TCT exercises and might therefore predict clinical outcomes after a full course of treatment. This study aimed to determine whether initial malleability of MMN and P3a to 1-h of auditory-based TCT predicts improvements in verbal learning and clinical symptom reduction following a full (30-h) course of TCT. Treatment refractory patients diagnosed with schizophrenia were randomly assigned to receive treatment-as-usual (TAU; n = 22) or TAU augmented with TCT (n = 23). Results indicated that malleability (i.e., change from baseline after the initial 1-h dose of TCT) of MMN and P3a predicted improvements in verbal learning as well as decreases in the severity of positive symptoms. Examination of MMN and P3a malleability in patients after their first dose of TCT can be used to predict clinical response to a full course of treatment and shows promise for future biomarker-informed treatment assignment.