This paper introduces a novel method to detect AI-generated fake face images or videos, known as Deep Fakes. The method leverages the inherent inconsistencies in the synthesis process, where Deep Fakes are created by splicing synthesized face regions into original images, leading to errors in 3D head pose estimation. The authors observe that these errors can be revealed by comparing head poses estimated from the central face region and the entire face. They develop an SVM classifier using this cue to differentiate between real and Deep Fake images. Experiments on real and Deep Fake datasets demonstrate the effectiveness of the method, achieving high AUROC scores. The paper also includes an ablation study to validate the importance of different features used in the classifier.This paper introduces a novel method to detect AI-generated fake face images or videos, known as Deep Fakes. The method leverages the inherent inconsistencies in the synthesis process, where Deep Fakes are created by splicing synthesized face regions into original images, leading to errors in 3D head pose estimation. The authors observe that these errors can be revealed by comparing head poses estimated from the central face region and the entire face. They develop an SVM classifier using this cue to differentiate between real and Deep Fake images. Experiments on real and Deep Fake datasets demonstrate the effectiveness of the method, achieving high AUROC scores. The paper also includes an ablation study to validate the importance of different features used in the classifier.