Media bias has long been a subject of scholarly interest due to its potential to shape public perceptions and behaviors. This systematic review leverages advances in natural language processing (NLP) to explore automated methods to detect media bias, addressing five core questions: it examines the definitions and operationalization of media bias, explores the NLP tasks addressed for its detection, the technologies used, and their respective outcomes and applied findings. This review also examines the practical applications of these methodologies and assesses the patterns, implications, and limitations associated with using artificial intelligence for media bias detection. Analyzing peer-reviewed articles from 2019 to 2023, the review initially identified 519 articles, which ultimately included 28 relevant ones. Significant heterogeneity is observed in bias definitions, affecting the analysis and detection approaches. The review highlights the predominant use of some methods and identifies challenges such as inconsistencies in problem definition, outcome measurement, and comparative method evaluation. Regardless of the conceptualizations of bias and the methods used, studies consistently identify bias in media outlets. Thus, studying media bias remains necessary for raising awareness and detection, and NLP methods are significant allies in this endeavor. This research aims to consolidate the foundations of recent advances in NLP for bias detection, encouraging researchers to focus on developing transparent, task-specific tools and work toward a consensus on a technical definition of bias and standardized metrics for its evaluation.