International Journal of Scientific Research and Engineering Development

Exploration on automated fake news discovery is desperately demanded, as the fleetly expanding digital and social media platforms have made the global problem of intimation indeed more burning. With an emphasis on their infrastructures, datasets, and performance comparisons, this thorough review analyzes and synthesizes deep literacy ways used to descry false information. The review illustrates the shift from conventional machine literacy ways to sophisticated neural models like CNN, RNN, LSTM, GRU, Transformer, and BERT- grounded systems by examining ten significant studies released between 2021 and 2025. The results show that multimodal and cold-blooded models that incorporate textbook, visual, and behavioral data attain much advanced delicacy rates frequently surpassing 95 percent with Motor- grounded models routinely outperforming earlier styles in terms of comprehending environment and expressing features. Nonetheless, patient obstacles like bias in datasets, conception across disciplines, multilingualism, and computational limitations still circumscribe wide relinquishment. In addition to outlining pivotal paths for developing secure and flexible automated systems for relating false information, this compendium offers a thorough review of slice- edge deep literacy ways for relating fake news.