This study presents an innovative approach to evaluating media representations of gender-based violence by integrating Natural Language Processing (NLP) techniques with the advanced capabilities of GPT-4, an Artificial Intelligence (AI)-based large language model. We developed a set of 27 expert-defined criteria to analyze a corpus of news articles, initially utilizing NLP methods for foundational text analysis. For more complex criteria, we employed GPT-4 and further enhanced its precision with fine-tuning. Our results indicate a significant increase in accuracy, achieving an overall 76% accuracy rate in content evaluation, which is 9% points higher than using NLP alone. This research introduces a novel media content analysis framework and paves the way for future enhancements in automated journalism assessment and ethical reporting.