This study addresses the pressing issue of detecting and recognizing deepfakes in the digital age. The authors collected a dataset of 83,000 images, including both pristine and deepfake images generated by nine GAN architectures and four diffusion models (DMs). They proposed a hierarchical multi-level approach to improve deepfake detection and recognition. The approach consists of three levels: (1) classifying real images from AI-generated ones, (2) distinguishing between images generated by GANs and DMs, and (3) recognizing the specific GAN and DM architectures used to generate the synthetic data. The experimental results demonstrated that the proposed method achieved over 97% classification accuracy, outperforming existing state-of-the-art methods. The models were also robust to various attacks such as JPEG compression and resizing, making them suitable for real-world applications like multimedia data analysis and forensic investigations. The hierarchical approach not only enhances the performance of deepfake detection systems but also aids in combating the spread of fake images and safeguarding the authenticity of digital media.This study addresses the pressing issue of detecting and recognizing deepfakes in the digital age. The authors collected a dataset of 83,000 images, including both pristine and deepfake images generated by nine GAN architectures and four diffusion models (DMs). They proposed a hierarchical multi-level approach to improve deepfake detection and recognition. The approach consists of three levels: (1) classifying real images from AI-generated ones, (2) distinguishing between images generated by GANs and DMs, and (3) recognizing the specific GAN and DM architectures used to generate the synthetic data. The experimental results demonstrated that the proposed method achieved over 97% classification accuracy, outperforming existing state-of-the-art methods. The models were also robust to various attacks such as JPEG compression and resizing, making them suitable for real-world applications like multimedia data analysis and forensic investigations. The hierarchical approach not only enhances the performance of deepfake detection systems but also aids in combating the spread of fake images and safeguarding the authenticity of digital media.