False news is one of the biggest problems that have emanated from the social media due to the spread of fake news. Coarse-grain, black-listing methods used in previous studies for the detection of fake news are incapable of capturing deeper semantics of the text with maximal precision and expansiveness. Subsequently with the introduction of deep learning more choices have opened up for analyzing content analysis and also the contextual features that are linked with fake news. This paper offers a cyclopedic analysis of the recent approaches in deep learning in relation to detection of fake news with special emphasis to the combined use of text, image and metadata data types. CNNs, RNNs, and transformers are some of the approaches we explain, coupled with their strengths and weaknesses, while also exploring the benefits of multimodal and hybrid systems in increasing the detectors’ performance. The future work research objectives are as follows: Comparing key advances Countermeasures for complex and progressing misrepresentation designs Anticipating developments in the field for increasing the dependability and efficacy of available fake news detecting systems.