Autobiographical memory (AM) dysfunction has been proposed as a neurocognitive mechanism underlying the development and maintenance of depression. However, case–control neuroimaging studies investigating the neural correlates of AM in depression have yielded inconsistent findings. The present study utilized neuroimaging meta-analyses to identify robust neural markers of AM dysfunction in depression and characterize the associated behavioral and network-level mechanisms. A preregistered neuroimaging meta-analysis (https://osf.io/35xtf) was conducted, incorporating data from 341 patients with unipolar depression, 82 individuals at risk of depression, and 261 healthy controls across case–control functional magnetic resonance imaging studies examining AM processing. Meta-analytic network-level and behavioral decoding analyses were performed to aid interpretation of the findings. Compared with controls, the depression group displayed increased activation in the right paracingulate cortex (dorsal anterior cingulate [dACC]) and precuneus, and decreased activation in the anterior insula during AM recall. Exploratory valence-specific analyses revealed that negative AM recall was associated with increased activity the dACC and precuneus. Meta-analytic decoding linked the dACC to the salience network and to domains related to negative affect and executive control, while the precuneus was associated with the default mode network and with processes related to social cognition and AM. Findings do not support prevailing models emphasizing altered amygdala and hippocampal function in AM deficits in depression. Instead, they highlight the involvement of core regions within the salience and default mode networks as key neural substrates of AM dysfunction. These regions may contribute to affective, social-cognitive, and mnemonic disturbances that shape the valence-specific nature of AM deficits in depression.