How to spot AI images: A fact-checker's guide - Snopes.com

What the report says

Snopes published a July 18, 2026 guide by Jack Izzo explaining how its fact-checkers evaluate whether viral images may have been generated by artificial intelligence. The article says improvements in generative AI have made image verification harder, but argues that users can still look for recurring warning signs rather than relying on a single clue.

The guide emphasizes visual details tied to light and camera behavior. Snopes says inconsistent or missing shadows can indicate a problem, because nearby objects should generally cast shadows in related directions when lit by the same source. It also points to unnatural depth of field, such as images that look uniformly “airbrushed,” place the wrong background elements in focus, or show distant and close objects with a level of clarity that would be unusual without editing.

Snopes also highlights shapes and small details that AI systems often mishandle. The article advises checking hands, ears, hair, body parts, wires and other complex elements for extra fingers, incomplete forms or objects that merge illogically. It says text remains useful to inspect, even though AI has improved, because generated images may still contain grammatical oddities, warped letters, incorrect logos or garbled official seals.

The publisher frames the tips as a checklist, not a definitive test: one flaw does not prove an image is fake, and some techniques may become less reliable as models advance. The broader significance is that AI-generated pictures are increasingly common online, making basic image-literacy skills more important for readers trying to avoid being misled by viral posts.