Two provenance signals can help, but neither can prove an image is real. Here is what each one tells you and where it falls short.
By FactInfoLive · Published 2026-11-12 · Updated 2026-11-12
The short answer
There is no single test that proves an image is real. Two signals can give positive evidence that an image came from AI or was edited: Google's SynthID watermark and C2PA Content Credentials. If neither is found, that tells you little.
SynthID
SynthID is an invisible watermark that Google adds to content made with its AI tools. Google says users can upload an image to the Gemini app and ask whether it is AI-generated. A report on the feature notes that it only works for images made with Google's AI, so Gemini cannot say whether another platform made an image. A security-company guide adds that SynthID now also covers images made with ChatGPT through a partnership announced in 2026, so check current coverage before relying on it.
Content Credentials (C2PA)
C2PA is a standard from the Coalition for Content Provenance and Authenticity. It attaches a signed record to a file that describes how it was created or edited. Popular Science reports it is supported by companies including OpenAI, Adobe and Google. You can inspect a credential with the official Content Credentials verify tool at contentcredentials.org/verify.
Why a negative result proves little
Images made with open-source models often carry no watermark at all, according to one security writer. Metadata can also be lost: a blog guide notes that a screenshot removes a file's metadata, including Content Credentials. And a credential is provenance, not a truth certificate: you still need to judge whether the signer or source is trustworthy.
A practical order of checks
1. Ask Gemini or use Google's SynthID Detector if you suspect Google-made AI.
2. Drop the file into the Content Credentials verify site and see if it carries a signed history.
3. Look for the platform's own AI label.
4. Run a reverse image search to find the earliest copy and the original source (see our reverse image search guide).
5. Only then look at the picture itself for odd details, and treat that as weak evidence.
Sources
Digital Camera World, on Gemini and SynthID: https://www.digitalcameraworld.com/tech/artificial-intelligence/googles-ai-will-now-both-create-fake-images-and-help-detect-them-as-gemini-gains-the-ability-to-detect-some-ai-generated-images
Popular Science, "How to really spot AI-generated images, with Google's help": https://www.popsci.com/technology/how-to-spot-ai-generated-images/
Malwarebytes, "How to tell if an image is AI-generated": https://www.malwarebytes.com/blog/news/how-to-tell-if-an-image-is-ai-generated
Malwarebytes, SynthID coverage update (2026): https://www.malwarebytes.com/?p=433207
Memx, "How to Spot an AI-Generated Image in 2026": https://memx.app/blog/how-to-spot-ai-generated-image/
Alejandro Rioja, "How to Tell If an Image Is AI-Generated: 6 Checks": https://alejandrorioja.com/how-to-tell-if-an-image-is-ai-generated/
Macmyths, on Content Credentials as provenance, not proof: https://macmyths.com/how-to-check-whether-images-on-a-website-are-ai-generated/