While artificial intelligence (AI) offers transformative potential for journalism, its application in newsrooms presents complex ethical challenges, particularly concerning bias in analyzing unstructured text data like reader comments. This study proposes an ethical newsroom clustering framework—comprising bias-aware preprocessing, model tuning and selection, and optional consensus fusion—to mitigate such biases. It introduces a robust median silhouette score for fairness-centric and trustworthy model selection. Experiments on both synthetic data and real-world New York Times comment datasets validated the framework’s efficacy, guided by the median silhouette, in preserving minority viewpoints and identifying nuanced subgroups that existing approaches obscured, thus avoiding common marginalization pitfalls. The proposed median-based silhouette score proved a more reliable and fair metric for evaluating clustering quality in noisy, real-world journalistic contexts. To the best of our knowledge, this is the first work dedicated to ethical newsroom clustering in journalism.