The detection of copy-move forgery has gained critical importance, particularly as digital manipulation techniques become increasingly accessible and sophisticated. In order to conceal or falsify visual information, copy-move forgery entails copying a portion of an image and placing it inside another image. Traditional approaches for detecting such forgeries face limitations when it comes to complex, high-resolution images or when minimal post-processing techniques are used to blend the duplicated sections. In order to accomplish reliable and effective detection of copy-move fraud, this paper suggests a unique framework that combines generative adversarial networks (GANs) with extended deep learning approaches, enhanced by textual feature extraction. Our approach leverages GANs to generate forged images and train the model to recognize subtle inconsistencies between authentic and tampered image regions. The architecture incorporates a two-step mechanism: first, an encoder-decoder structure with a GAN-based discriminator to identify inconsistencies in pixel-level features, followed by a textual feature extraction layer that analyzes embedded metadata and any residual textual artifacts that could indicate forgery. This combined multi-modal approach allows for improved localization of copied regions and minimizes false positives by cross-referencing both visual and textual inconsistencies.