SSkrubly
enOpen a tool
← All articles

September 17, 2026 · 6 min read

How to Tell If an Image Was Made With Gemini (SynthID)

Check whether an image came from Google Gemini: SynthID watermark vs C2PA metadata, how to verify in the Gemini app, and what a metadata checker can show.

Checking an image for Google's SynthID watermark and metadata

Google's Gemini image models (including the ones nicknamed Nano Banana) can produce very realistic pictures, so it's reasonable to want to check where an image came from. There are two different signals involved, and they need different tools. Here's how each works and which one you can actually check.

Signal 1: SynthID, the invisible watermark

Google DeepMind's SynthID embeds a watermark directly into the pixels of images made or edited with Google's AI. You can't see it, and it's designed to survive common changes like cropping, compression and screenshots. Because it's in the pixels, not the file's metadata, no metadata tool (including Skrubly's) can detect it. Detection needs Google's own tools.

Signal 2: C2PA Content Credentials (metadata)

Google has also started adding C2PA Content Credentials to images from its newer image models in the Gemini app and some Google Cloud products. According to Google's documentation, these credentials are signed by Google and name the Google service that produced the file. Unlike SynthID, this information lives in the file's metadata, so it can be read by a metadata checker, and it can be lost when a file is screenshotted, re-saved or uploaded to a platform that strips metadata.

CheckWhat it findsSurvives screenshots/re-uploadsTool
SynthIDGoogle AI watermark in pixelsDesigned toGemini app verification
C2PA Content CredentialsSigned record of AI generationOften lostMetadata/C2PA checker
Visible watermarkGemini sparkle logo on some imagesEasy to crop outYour eyes
Visual cluesOdd text, hands, reflectionsYesCareful inspection

How to verify in Gemini

At the time of writing, Google's help centre describes a verification feature built into the Gemini app and gemini.google.com:

Google notes limits: the check only recognises Google AI content, there are daily usage limits, results can be unclear for small or lightly edited images, and screenshots should be cropped tightly around a single image. Google has said it plans to bring this kind of verification to more of its products, so the exact options may change.

Check the metadata with the AI Image Checker

If you have the original file, Skrubly's AI Image Checker reads C2PA Content Credentials and generator tags in its metadata. If a Gemini image still has its credentials, you'll see them. It runs in your browser, so the image isn't uploaded. What it can't do: detect SynthID, or tell you anything when metadata has been stripped. No markers means no markers, not proof the image is real.

Read an image's Content Credentials and AI generator tags.

Open the AI Image Checker

A sensible order of checks

More ways to judge whether a photo is real.

How to tell if an image is AI

Why no single check is final

Each signal answers a narrow question. SynthID tells you about Google AI and some partners, not every generator. C2PA tells you what a file says about itself, if the data is still there. Visual inspection is subjective. Put together with where the image came from, they give you a reasonable judgement, not certainty. Be especially careful before accusing someone of faking an image based on one tool's result.