Neuropsychiatric illnesses including mood disorders are accompanied by cognitive impairment, which impairs work capacity and quality of life. However, there is a lack of treatment options that would lead to solid and lasting improvement of cognition. This is partially due to the absence of valid and reliable neurocircuitry-based biomarkers for pro-cognitive effects. This systematic review therefore examined the most consistent neural underpinnings of cognitive impairment and cognitive improvement in unipolar and bipolar disorders. We identified 100 studies of the neuronal underpinnings of working memory and executive skills, learning and memory, attention, and implicit learning and 9 studies of the neuronal basis for cognitive improvements. Impairments across several cognitive domains were consistently accompanied by abnormal activity in dorsal prefrontal (PFC) cognitive control regions—with the direction of this activity depending on patients’ performance levels—and failure to suppress default mode network (DMN) activity. Candidate cognition treatments seemed to enhance task-related dorsal PFC and temporo-parietal activity when performance increases were observed, and to reduce their activity when performance levels were unchanged. These treatments also attenuated DMN hyper-activity. In contrast, nonspecific cognitive improvement following symptom reduction was typically accompanied by decreased limbic reactivity and reversal of pre-treatment fronto-parietal hyper-activity. Together, the findings highlight some common neural correlates of cognitive impairments and cognitive improvements. Based on this evidence, studies are warranted to examine the reliability and predictive validity of target engagement in the identified neurocircuitries as a biomarker model of pro-cognitive effects.