Aimee Stanton

Researchers from universities in the UK, Australia and Canada show that people can significantly improve their ability to identify AI‑generated faces through targeted training. The study highlights perceptual cues such as symmetry, proportionality and expressiveness that help distinguish real images from deepfakes. Training raised participant accuracy from around 40% to 80%, with some nearing perfect identification. The article also outlines risks associated with deepfakes, including fraud and political espionage, while noting potential beneficial uses such as age‑progression imagery in missing‑person cases.