Recognising AI Generated Image Artifacts
Launch library · evergreen read
Generated images have distinctive weak points, even as the underlying technology keeps improving month after month. Hands with an unnatural number of fingers, text within the image that dissolves into meaningless shapes once you look closely, and jewellery or fabric patterns that do not quite repeat correctly remain common tells genuinely worth checking for carefully before trusting an image at face value.
Backgrounds are often a second, equally revealing weak point worth examining closely. Repeating textures that do not logically continue across the frame, reflections that do not match what should actually be casting them, and lighting that falls inconsistently across different parts of the very same scene are all worth a careful, deliberate second look before accepting an image as genuine.
These specific flaws are gradually improving out of existence over time as the technology matures further, so relying on visual artifacts alone is steadily becoming less reliable as a long term strategy. Checking a striking image's origin and its subsequent spread, alongside any visual scrutiny applied, remains the considerably sturdier habit to build for the future.