<p>Lung cancer (LungC) is one of the most critical causes of mortality worldwide, with intricate mechanisms such as genetic alterations and epigenetic modifications contributing to its onset and spread. The illness is frequently identified in progressed phases, restricting patients’ therapeutic choices. Large Language Models (LLMs), among the most sophisticated artificial intelligence (AI) innovations, have demonstrated remarkable potential in enhancing the identification and management of LungC. By concurrently analyzing textual and visual information, these models can uncover latent trends, forecast therapy reactions, and expedite treatment-related decision-making. Additionally, merging diverse data sources, such as medical imaging interpretations and genomic records, enhances the precision of early recognition and decreases the duration required for making treatment decisions. Nevertheless, employing LLMs presents several challenges, including the need for extensive computational capabilities, issues with data uniformity, and limitations in the clarity of decision-making processes. These concerns necessitate the formulation of more efficient methodologies and more resilient algorithms. This survey delivers an in-depth categorization of the uses and obstacles associated with leveraging LLMs for LungC diagnostics and treatment, alongside the most recent breakthroughs in this area. The findings of this survey indicate that LLMs have boosted the precision of early lung tumor identification (PELI) by 35% and the prediction rate of treatment effectiveness (PRTE) by 25%. Moreover, applying these models has enhanced the recognition speed of suspicious lesions (RSSL) by 22%. Ultimately, by analyzing the existing difficulties and suggesting remedies to tackle them, this survey will support academics and experts in this discipline in formulating novel approaches to refine LungC detection and therapy strategies, as well as achieving a more profound comprehension of LLMs’ role in the medical domain. </p>