AI-Generated Image Identification Using Explainable AI Techniques

The increasing availability of AI-generated images has made distinguishing synthetic content from real images a critical challenge in digital forensics, misinformation control, and content authenticity verification.This paper proposes an explainable deep learning-based approach for identifying AI-generated images using a convolutional neural network built on a pre-trained ResNet50 architecture with transfer learning.The model learns discriminative visual features to classify images as real or AI-generated with high accuracy.To enhance transparency and user trust, Grad-CAM-based Explainable AI is integrated to visually highlight the image regions that influence the model's predictions.Additionally, a confidence-based uncertainty mechanism is employed to flag low-confidence predictions and recommend manual verification, reducing the risk of incorrect automated decisions.Experimental results on diverse datasets demonstrate strong classification performance and generalization across different image sources.The proposed system effectively combines accuracy, interpretability, and reliability, making it suitable for real-world applications such as deepfake detection, digital content verification, and forensic image analysis.