Google marks AI images in several layers. Some are metadata you can remove; one — SynthID — is in the pixels.
🔒 Your files never leave your device — processing happens in your browser.
The layers Google uses
- Visible mark — Gemini app images can show a small sparkle logo in a corner. That is part of the picture, not metadata.
- SynthID — an invisible watermark woven into the pixels by Google DeepMind. It survives cropping and compression and is checked by Google’s own tools.
- Metadata — depending on the product and date, files can carry IPTC/XMP “AI-generated” tags, C2PA Content Credentials, the generator name and an
AI-style file name such as
Gemini_Generated_Image_xxxx.png.
What Clean Image Metadata removes
All metadata layers: C2PA, IPTC/XMP, EXIF and text chunks — losslessly. Use Neutral names to drop the Gemini file name. Check the result with the checker.
What it does not remove
SynthID and the visible logo are part of the image itself. Removing them would mean altering the pixels, which is not what a metadata tool does — and Google designed SynthID precisely so AI images can still be identified. If you post Gemini images where AI content must be disclosed, disclose it.
Frequently asked questions
Can metadata removal remove SynthID?
No. SynthID is not metadata; it lives in the pixel values.
Why is my Gemini image still detected as AI after cleaning?
Platforms can use SynthID or their own visual detectors in addition to metadata.