The internet has so much information that it can be overwhelming. With a myriad of articles on any subject under the sun, from politics to sports it leaves us in a daze as to how reliable they are. Existing research has tackled truth discovery, but there is still a gap: finding trustworthy news gathered from multiple sources with fewer errors and shorter retrieval time. This paper develops an innovative system that utilizes natural language processing techniques for a holistic determination of the reliability of online news. The proposed method initially gathers data from official as well as unofficial news sources followed by preprocessing with tokenization and generation of embeddings using Bidirectional Encoder Representations from Transformers (BERT) model to encode the semantic content of the article. The cosine similarity between the BERT embed- dings of the official and unofficial articles assesses the truthfulness. This is further fine-tuned by a consensus scoring mechanism that applies pairwise cosine similarity across embeddings from all the unofficial articles and delivers an aggregated consistency metric. Sentiment analysis is incorporated by employing the BERT-based TextBlob and the Multilingual BERT sentiment model to classify every tone in the article. Political bias detection is delivered using a multilingual DeBERTa model which captures the ability of words to contain or show up with potential slants according to ideological or linguistic markers. The results of experiments prove mean deviations of the truth value to be 0.17 and the scores above 0.8 relevance score for the most credible sources, thus, pointing to the efficiency of such a system in real-time credibility and multi-dimensional assessments. Ultimately, this work aims to empower users with a robust tool for navigating the online news landscape, ensuring they can access trustworthy information swiftly and effectively.