With the growing use of digital images in online communication and social media, verifying whether an image is authentic has become increasingly important. Image manipulation techniques such as copy-move, splicing, inpainting, and object removal can change the meaning of visual information. At the same time, JPEG compression and repeated recompression can weaken the forensic traces that are useful for detection. This review looks at two AI-based approaches for detecting digital image forgery. The first study combines Error Level Analysis (ELA) with the VGG-16 convolutional neural network to classify images as original or tampered. The second study focuses on compressed images shared through social networks and uses a modified ShuffleNet V2 architecture for lightweight forgery classification. The two studies are reviewed based on their objectives, datasets, preprocessing or feature representation, deep learning architecture, compression conditions, findings, and limitations. The comparison shows that ELA with VGG-16 can help expose compression-related manipulation artifacts, while the modified ShuffleNet V2 approach focuses on the practical challenge of detecting forgery after JPEG and social-network recompression. The review highlights dataset diversity, compression robustness, computational efficiency, and generalization across different manipulation and sharing conditions as important research gaps.

Keywords— Artificial Intelligence, Digital Image Forgery, Deep Learning, Error Level Analysis, VGG-16,

ShuffleNet V2, Image Compression.