Is this image AI? What you can check before sharing.

A detector may confirm one known AI watermark. It cannot certify every other image as genuine.

A dramatic photo lands in your group chat with one line: “This just happened.” Before you pass it on, check where it came from. A convincing picture can be old, cropped, edited or generated—and the pixels alone may not settle which.

Start with the source. Search a distinctive phrase from the caption, look for the earliest post you can find, and compare what reliable outlets or the people involved say. Reverse-image search can uncover older copies or a different context. It cannot, by itself, prove when or how an image was made.

Look for provenance information. Some images include Content Credentials that record origin or edits. If they are present, inspect the details and who signed them. Many genuine images have no credentials, and credentials can be missing after a file is copied or transformed. Their absence is not proof of fakery.

For Google AI, there is a narrower check. Google says Gemini can check images and videos for its SynthID watermark. A detected watermark is evidence that Google’s generative AI was used in the marked content. The check does not identify media made by every other AI system. No SynthID result does not mean “definitely a camera photo.”

Before sharing: verify the earliest source and the claim’s context. Treat a watermark result as one clue, never an all-purpose real-or-fake verdict.

Availability and supported media vary by service. These checks cannot establish the full history of every image.

Sources: Google, image verification in Gemini with SynthID (2025); Content Credentials, verification tool.