Built a deep learning model to classify images as AI-generated or human-captured. Used DenseNet121 for feature extraction, achieving 94.6% accuracy. Included Grad-CAM for explainability. Supports digital forensics and fake media detection in the age of generative AI.