Abstract: Ground-glass nodules (GGNs) are common imaging manifestation in the early screening of lung adenocarcinoma. With the widespread use of low-dose computed tomography (LDCT) in lung cancer screening, the detection rate of GGNs has significantly increased. According to the presence or absence of a solid component, GGNs are mainly classified into pure ground-glass nodules (pGGNs) and mixed ground-glass nodules (mGGNs), which differ in their natural course and biological behavior. In general, pGGNs tend to progress more slowly, whereas mGGNs are more likely to develop invasive features. The vast majority of pGGNs remain stable for years, but some pGGNs and mGGNs may show an increase in size or in the solid component. The "indolence" observed on the surface of GGNs hides complex genomic, metabolic, and immune changes, which are difficult to capture with traditional image-based data. In recent years, multi-omics analysis, radiomics, and artificial intelligence models have provided new tools for identifying high-risk GGNs. However, there is still controversy over the clinical generalizability, interpretability, and standardization of these models. Furthermore, there is no consensus on whether surgical resection is required. This article reviews the molecular mechanisms, metabolic, and immune microenvironment changes involved in the progression of GGNs, discusses the advantages and limitations of imaging prediction models, and combines domestic and international guidelines and survival studies to explore the controversial points in follow-up and surgical strategies, aiming to provide references for the personalized management of GGNs.