TLDR: New research from a collaboration of UK universities reveals that just five minutes of targeted training can substantially enhance an individual’s capacity to distinguish between real human faces and those created by advanced artificial intelligence, offering a promising tool in the fight against deepfakes and online misinformation.
A groundbreaking study conducted by scientists from the University of Reading, the University of Greenwich, the University of Leeds, and the University of Lincoln has demonstrated that a mere five minutes of training can dramatically improve a person’s ability to identify faces generated by artificial intelligence. This timely discovery comes as AI-created imagery becomes increasingly sophisticated and realistic, posing significant challenges to digital authenticity and security.
The research, which involved 664 participants, tested their proficiency in differentiating between genuine human faces and those produced by StyleGAN3, a highly advanced computer software known for generating remarkably lifelike images. Initially, without any specific training, participants struggled to identify the synthetic faces. ‘Super-recognisers,’ individuals with exceptional face recognition skills, correctly identified fake faces only 41% of the time, while participants with average recognition abilities managed a mere 31%. For context, random guessing would yield an accuracy of approximately 50%, highlighting the inherent difficulty in spotting these AI creations.
The brief training intervention proved to be remarkably effective. Participants were educated on common rendering errors and subtle visual cues often present in AI-generated images, such as unusual hair patterns, inconsistent numbers of teeth, unnatural eye shapes, irregularities in facial symmetry, and discrepancies in lighting or skin textures. Following this short session, detection rates soared. Super-recognisers achieved an impressive 64% accuracy in identifying fake faces, and typical participants saw their accuracy rise to a commendable 51%.
Dr. Katie Gray, the lead researcher from the University of Reading, emphasized the critical implications of these findings. “Computer-generated faces pose genuine security risks. They have been used to create fake social media profiles, bypass identity verification systems and create false documents,” Dr. Gray stated. She further noted a concerning trend where “People often judge AI-generated faces as more realistic than actual human faces.” The study’s results suggest that this brief and easily implementable training procedure could be a vital tool in addressing these real-world problems, particularly in online identity verification.
Also Read:
- Artificial Intelligence Revolutionizes Educator Development and Personalized Learning, New Studies Reveal
- Public Trust in Artificial Intelligence Outstrips Government and Social Media in New UK Study
The study also observed that the training benefited both super-recognisers and average participants equally. This suggests that super-recognisers might employ different visual cues than typical observers when discerning synthetic faces, rather than simply being better at spotting rendering errors. The findings underscore the potential for quick, accessible educational modules to serve as a scalable defense against the growing threat of synthetic media, combating misinformation, fraud, and online deception. The research was published on November 12, 2025, in Royal Society Open Science.
