New research suggests folks are relying on 'outdated visual cues' to identify AI-generated faces—but I did get 14/20 on the test
PC Gamer
by Jess Kinghorn February 19, 2026
AI-Generated Deep Dive Summary
New research from UNSW Sydney reveals that most people struggle to distinguish between AI-generated faces and real ones, performing only slightly better than chance in identification tests. The study involved 125 participants, including "super-recognizers" who excel at face recognition tasks, but even they showed limited success in identifying AI faces. While super-recognizers performed marginally better than average participants, the overall results highlight a persistent challenge in detecting synthetic faces.
The research attributes this difficulty to the fact that many generative AI models produce "hyper-average" appearances, which are statistically likely and thus harder for humans to spot. Participants often relied on outdated visual cues like distorted teeth or mismatched accessories, which modern AI models have largely overcome. The study underscores how advancements in AI have made it increasingly difficult to rely on traditional tell-tale signs.
This development has significant implications for gaming and virtual interactions, where realistic avatars and synthetic identities are becoming more common. While context clues—such as inconsistent social media behavior or repetitive content—can help identify potential AI-generated personas, the study suggests that even sophisticated users may struggle to discern real from fake faces in many cases.
Ultimately, while AI-generated faces remain somewhat detectable through subtle cues like "average" appearances, the research emphasizes the need for improved methods of identifying synthetic identities. For gamers and tech enthusiasts, this raises questions about trust, privacy, and the potential for misuse in virtual environments. As AI continues to evolve, distinguishing between real and fake faces may become even more challenging, requiring innovative solutions beyond relying on visual cues alone.
Verticals
gamingpc
Originally published on PC Gamer on 2/19/2026