Major Depressive Disorder (MDD) is a globally prevalent mental health condition with high disability rates. Its diagnosis currently relies primarily on clinical interviews and scale assessments, and there remains a lack of reliable objective biomarkers. Event-related potential (ERP), as a non-invasive, real-time brain function detection technology with millisecond-level temporal resolution, can directly reflect neural activities during cognitive and emotional tasks, providing a new perspective for the diagnosis and treatment of MDD. This review systematically summarizes key findings from ERP studies on cognitive and emotional processing abnormalities in MDD, indicating that multiple ERP components exhibit characteristic alterations in MDD patients, reflecting dysfunctions in cognitive control, emotion regulation, and reward processing across multiple stages. Although these indicators commonly demonstrate transdiagnostic heterogeneity and state-dependent variability, their dynamic patterns and multi-component integrated analyses show potential for assisting in diagnosis, predicting treatment efficacy, and monitoring interventions including pharmacotherapy, neuromodulation, and psychological therapy. Current research still faces challenges such as insufficient paradigm standardization, clinical heterogeneity, and reproducibility issues. Future efforts should focus on promoting multi-center collaboration, integrating multimodal data, and employing machine learning modeling to advance the translation of ERPs from experimental markers to clinically practical tools, ultimately contributing to personalized diagnosis and treatment of MDD.