AI-based Multi model Misinformation Detection using NLP and CNN Models

The increasing problem of multimodal fake news is that which takes the form of disinformational text and reworking or out-of-context pictures, has made the old-fashioned text based one-dimensional methodologies of detection becoming less adequate. In order to manage this issue, the Cross-Check Fusion Model (CCFM) is suggested as an Artificial Intelligence-based system that combines transformer-based Natural Language Processing (BERT) to analyze text and Convolutional Neural Networks (CNN) to process the images. The cross-modal attention is used to combine textual and visual information and a fact-checking layer that uses external verification systems like the Google FactCheck API to determine the truthfulness of the facts. The model is trained and tested on multimodal benchmark FakeNewsNet multimodal datasets (GossipCop and PolitiFact) which consist of real and fake news pairs. The experimental findings confirm that CCFM shows 85.14 accuracy which is better than the unimodal baselines. This method does not only improve the accuracy of detection, but also the reliability and transparency of misinformation detection, and provides a scalable and reliable solution to digital media and journalism.