No single visual flaw, detector score, or metadata field can settle how an image was made. Start with the image’s source, then compare several kinds of evidence.
Inspect the image at more than one scale
Look first at the scene as a whole. Do the shadows, reflections, perspective, and object relationships make sense together? Then zoom in on areas with small details such as hands, signs, jewelry, or repeating patterns. Generated images can contain inconsistencies there, though current systems often produce convincing detail and ordinary photographs can have artifacts too.
Text can be misspelled or arranged oddly. Reflections may not match the objects in front of them. Repeated windows, leaves, or buttons may vary in a patterned way. These are leads for closer inspection, not a reliable checklist of generator flaws.
Use an AI classifier as one signal
An image detector compares the file with patterns learned by a model. Motchilllz’s AI Analyzer returns a score and a label such as “Possibly AI-generated.” The score describes that model’s output. It does not identify a creator, prove that a photo is authentic, or establish that an image was generated.
Edited photos, screenshots, compression, and image sources outside a model’s training examples can change the result. A low score does not rule out generation, and a high score can be a false positive. Read why AI image detectors can be wrong before relying on a result.
Check metadata and provenance
An EXIF Detector can show camera and capture fields when they remain in the file. Those fields can be removed or edited. Their absence is common after screenshots, exports, and social sharing, so it is not evidence of AI generation. Their presence also does not prove that a scene is genuine.
Look for a source page, a creator’s explanation, or verifiable Content Credentials when available. Provenance can provide a record of a file’s history, but it does not establish that the depicted event is true.
Compare evidence before deciding
Save the original file when possible. Find where it first appeared, compare versions, and check whether the surrounding account or publication gives a traceable source. Treat visual clues, model scores, and metadata as separate observations. When they disagree or the source is unknown, the honest answer may be that the image’s origin is unclear.
Use the image authenticity checklist to work through those checks in order.