Can you remove the Gemini watermark?
Gemini images carry three marks: a visible sparkle, a C2PA record and an invisible SynthID watermark. Which ones you control, and which ones stay.
What are the watermarks on a Gemini image?
Gemini, including its Nano Banana image model, marks images in three separate layers. Each one lives in a different place, so each one behaves differently.
| Mark | Where it lives | Who can see it | Can you turn it off? |
|---|---|---|---|
| Visible sparkle | Drawn on the image | Anyone looking at it | Yes, in Gemini’s settings |
| Content Credentials | Signed metadata in the file | Any C2PA reader | It disappears when metadata goes |
| SynthID | Spread through the pixels | Google’s detector only | No |
Google’s help page for the watermark setting lists all three: visible watermarks, SynthID on all media created or edited with Gemini Apps, and Content Credentials metadata describing the content’s origin.
Can you turn off the visible Gemini watermark?
Yes, through Gemini’s own settings. On gemini.google.com, open Settings, then Media Watermark, and switch it off. The same option exists in the Android and iPhone apps. It applies to images you generate afterwards, not to images already made.
Google notes two limits. In India, South Korea and Vietnam the setting only appears for AI Ultra subscribers, and everyone else there always gets the visible watermark. And the switch only controls the visible mark: “It doesn’t affect SynthID watermarks or Content Credentials,” in Google’s words.
The Gemini image we downloaded for our tests on 27 September 2026 had no visible sparkle in its corner. It still carried the other two marks.
What happens to the C2PA record?
That image’s metadata held a C2PA manifest signed by Google’s C2PA library, with two recorded actions: “Created by Google Generative AI.” and “Applied imperceptible SynthID watermark.” The full field list is in AI image metadata.
Because it is metadata, the record is fragile. A screenshot drops it, most chat apps and social networks re-save the file without it, and any metadata cleaner can delete it. We ran our own Remove AI traces option on the file: the manifest was gone and our checker found no signal afterwards.
Can SynthID be removed?
Not by removing metadata, and not by any setting. SynthID is an invisible pattern that Google DeepMind embeds in the pixels while the image is generated, and Google designed it to survive common edits such as cropping, filters and compression.
Our test shows the practical side. After we removed the C2PA record, we compared the decoded pixels of the two files and found them identical. Whatever SynthID put into the image was still there, and the Gemini app can still recognise the picture as made with Google AI. We explain how the watermark works in what is SynthID.
What can you do with a Gemini image, then?
If you do not want the sparkle on your own images, use Google’s setting before you generate them. If you want a file without the C2PA record, for privacy or to keep a file small, removing metadata does that and leaves the picture exactly as it was. If you need to know whether someone else’s image came from Gemini, use both checks: our AI image detector for the Content Credentials, and the Gemini app for SynthID.