You usually cannot identify a ChatGPT image from one visual feature or metadata tag. Generation and editing workflows change, and an image may be resized or modified before you see it.
Start with the source
Check who shared the image and whether the post explains how it was made. An original conversation, source page, or file from the creator can provide context. Reposts may lose both captions and original metadata.
Inspect clues, then run a classifier
Look for inconsistent geometry, lettering, reflections, and object continuity. Treat those observations as leads. Generated images can avoid older artifacts, while photography and retouching can produce similar irregularities.
Upload the image to the AI Analyzer for a probabilistic score. It estimates whether the file resembles the model’s AI class. It does not determine that ChatGPT specifically produced the image. A strong score is not proof, and a low score does not rule generation out.
Check metadata and provenance
Review any software, date, camera, or other EXIF fields with the EXIF Detector. Metadata can be stripped by sharing or export workflows and can be edited. Do not assume that every image generated through ChatGPT carries the same metadata.
Some files may include signed Content Credentials describing provenance. If a credential is available, verify it with a tool that supports that standard and read what the credential actually says. Motchilllz does not currently validate Content Credentials.
When you cannot find the original source or a verifiable record, describe the result as uncertain. See the image authenticity checklist for other checks.