The rapid development in deep learning-based generative softwares and image rendering tools has led to generation of massive photorealistic digital content – Fake images, fake videos, fake speech, etc. Such fake digital media may result in communication of misinformation, forgery of digital data, and losing trustworthiness in the information source. This poses a significant challenge to the field of digital forensics’ techniques, to which our present work attempts to make a contribution, by addressing the problem of differentiating AI-generated images from real photographs, using transfer learning and multi-branch fusion model. We propose a multi-branch model that integrates two pre-trained Vision Transformer models (DINO (self-distillation with no labels) and Contrastive Language – Image Pretraining (CLIP)) to extract complementary global features, along with a forensic and a hand-crafted feature branch, which extract low-level discriminating cues. These features are complimentary to each other and hence contribute in improving the robustness and performance of the model. These features from the four branches are adaptively weighted and combined by a cross-attention module, to give a fused and rich embedding. The model is further optimized by using augmentation-invariant loss, center loss and supervised contrastive loss in addition to the cross-entropy loss function. This framework achieves improved accuracy of 95.83% on PRCG dataset, 96.79% on CIFAKE dataset and 99.15% on GenImage dataset as compared to baselines. It also achieved stable cross-generator performance and enhanced robustness against real world corruptions like Blur, Noise, Compression, and others. The experimental results show a good separability between the classes, and enhanced performance on publicly available datasets.