Deepfakes use AI and deep learning algorithms to create digitally manipulated images, videos or audio recordings that can alter a person’s appearance or voice, making them increasingly convincing and difficult to detect. As computational power and technology advance, these manipulations are becoming more sophisticated, posing significant challenges in the fight against disinformation and defamation. Currently, deepfake detection is largely reliant on AI and machine learning models. However, metadata—information that describes the content and technical specifications of an image—can serve as a crucial artefact in forensic analysis. Metadata is embedded within image files, often generated automatically during capture or creation, and can get altered during the process of creating a deepfake. This study examines the differences in metadata between original images and their deepfake counterparts to assess the effectiveness of metadata analysis in detecting manipulated images from a forensic perspective. We analysed 100 original images and their corresponding deepfakes to identify metadata alterations and evaluate the potential of this approach in deepfake detection.