Visual clues can help you decide where to look more closely. They do not work as a reliable pass-or-fail test. Generators change, and ordinary images can contain strange details after motion blur, retouching, or resizing.
Details to inspect
- Text and labels: zoom in on small signs, packaging, or lettering. Some generated images contain malformed words or letter shapes. A low-resolution photograph can produce a similar result.
- Geometry: compare repeated objects such as windows, rails, or tiles. Look for edges that bend, end abruptly, or change spacing without a scene-based reason.
- Reflections and shadows: check whether a reflection follows the object’s shape and viewpoint. Compare shadows with the scene’s apparent light sources.
- Anatomy and object continuity: inspect how fingers hold an object, how straps meet clothing, or how frames and handles connect. Cropping can hide a connection, so include surrounding context.
- Repeating texture: look at hair, foliage, fabric, and background details. Unnatural repetition or a sudden shift in texture can be a clue, but compression can make patterns too.
These observations become weaker when you see a small crop. Open the largest available copy and compare it with another version when possible. Do not inspect only the area that looks suspicious; check whether the overall scene remains coherent.
Combine visual review with other evidence
Modern image generators often avoid artifacts that appeared in older examples. A clean hand or legible sign does not rule out generation. An artifact does not prove generation either.
The AI Analyzer provides another probabilistic signal. EXIF metadata and source history may add context, but metadata can be stripped and an image can be edited after capture. Use the clues together and keep their limits visible. For a worked process, see how to tell if an image is AI-generated.