This article presents a comparative analysis of two methodologies for Fake News detection based on YIX and Our proposed model (AI). The YIX methodology integrates YOLO v3 for object detection, InceptionResNetV2 for feature extraction, and XGBoost for classification. In contrast, the AI (InceptionResNetV2 AdaBoost) methodology enhances the YIX approach by eliminating YOLO v3 to mitigate potential feature concealment issues, incorporating modified InceptionResNetV2 architecture with additional dropout and dense layers, and substituting XGBoost with AdaBoost for improved binary classification performance. The evaluation is conducted on two datasets: CatsandDogs, comprising images of two animal types, and Caleb-DF, containing over 20,000 images classified into real and fake classes. Each dataset consists of 5,000 images for testing purposes. Results demonstrate that the AI methodology yields superior performance compared to YIX. For the CatsandDogs dataset, ours achieves an accuracy of 0.99, specificity of 1, and recall of 0.98, whereas YIX achieves an accuracy of 0.94, specificity of 0.94, and recall of 0.94. Similarly, on the Celeb-DF dataset, AI achieves an accuracy of 0.91, specificity of 0.93, and recall of 0.88, surpassing YIX with an accuracy of 0.84, specificity of 0.85, and recall of 0.81. Notably, the AI model achieves 91% higher accuracy than YIX, indicating significant improvement in positive prediction accuracy. These findings underscore the effectiveness of the AI methodology in enhancing object detection and classification performance, particularly in scenarios involving complex image datasets. The results highlight the importance of methodological enhancements and algorithmic modifications in advancing the capabilities of computer vision systems.