What Gemini and Nano Banana images contain: C2PA, SynthID and EXIF
Short answer: Gemini image metadata can include up to three things. One is an invisible SynthID watermark, which is part of the pixels, not the metadata. Since November 2025, Nano Banana images also carry a signed C2PA record (Content Credentials) saying Google AI made them. Some Gemini PNGs also hold an EXIF block in a format many viewers can't read. Metadata tools can show and remove the C2PA record and EXIF block. SynthID stays.
The three layers in a Gemini image
| Layer | Where it lives | What it says | Survives metadata removal? |
|---|---|---|---|
| SynthID watermark | In the pixels | "Made or edited with Google AI" | Yes |
| C2PA Content Credentials | Metadata block in the file | Which tool made the image, signed by Google | No |
| EXIF | Metadata block in the file (PNG text chunk) | Technical fields; contents vary | No |
1. SynthID: the watermark in the pixels
Google describes SynthID as "an invisible digital watermark" added to AI-generated images and video. It isn't a metadata field. It is a pattern spread through the image itself, and Google says it is designed to withstand cropping, filters and lossy compression. Saving the image in another format, taking a screenshot or stripping the metadata does not remove it.
Google also says other companies are adopting it. In its I/O 2026 announcements (20 May 2026), it named OpenAI, Kakao and ElevenLabs as bringing SynthID to their own AI content. So a SynthID result won't always mean Gemini.
2. C2PA: the signed "made with AI" record
On 20 November 2025 Google announced that images made with Nano Banana Pro (Gemini 3 Pro Image) in the Gemini app, Vertex AI and Google Ads would have C2PA metadata embedded. Its Nano Banana 2 post (February 2026) describes pairing SynthID with "interoperable C2PA Content Credentials".
C2PA is the same standard OpenAI, Adobe and camera makers use: a manifest, signed with a certificate, that records which tool created or edited the file. Platforms such as Instagram read it when deciding whether to show an AI label. C2PA content credentials explained covers what a manifest holds.
Google's posts don't say whether every Gemini model, or images from Gemini's API, get the same record. Don't assume a Gemini image has C2PA. Check the file.
3. EXIF: present, but often unreadable
In August 2025 a developer reported on Google's AI developer forum that a PNG downloaded from Gemini stored its EXIF in a compressed text chunk labelled Raw profile type APP1, instead of the usual Raw profile type exif. exiv2, GIMP, macOS Preview and ACDSee all failed to decode it. A Google staff member said they would escalate it, and the thread shows no fix.
In practice, a viewer may say "no EXIF" or show an error for a Gemini PNG that does have an EXIF block. ExifTool reads PNG text chunks and is the best way to see what is really in there.
How to check a Gemini image
For SynthID, use Google's own tools. Only Google can check for SynthID. Upload the image to the Gemini app and ask "Was this created with Google AI?" Google says Search (Lens, AI Mode and Circle to Search) and Gemini in Chrome can run the same check. No third-party metadata viewer can see SynthID, because there is nothing in the metadata to see.
For C2PA and EXIF, look at the metadata.
- Keep the file exactly as Gemini saved it. A screenshot or a copy re-shared through a chat app has usually lost its metadata already.
- Drop it into the image metadata viewer. It runs in your browser, accepts JPEG, PNG and WebP, lists the readable metadata fields, and flags when C2PA markers are present. It detects C2PA but doesn't parse or verify the signature.
- To read the full manifest (signer, date, actions), use a C2PA verifier such as the Content Authenticity Initiative's Verify site or the open-source
c2patool. Our content credentials checker also detects whether the file has C2PA markers. - For everything else, run ExifTool:
exiftool -a -G1 gemini_image.png
-G1 shows which block each field came from, so you can tell the PNG text chunks apart from the C2PA data.
Since May 2026, Google says the Gemini app can also check C2PA Content Credentials, including ones from other companies' tools.
Removing the metadata, and what that does and doesn't change
There are honest reasons to clean the file before sharing. You may not want a signed record of your Google account's tool use in a family photo you edited with Gemini. Or you may just not want EXIF fields you can't read. The AI info tag remover redraws the image on a canvas in your browser and saves a new file. That drops EXIF, XMP, IPTC and C2PA together, and nothing is uploaded. JPEGs are re-saved at quality 0.95, so keep the original.
If the image is AI-generated, though, the C2PA record is accurate, and removing it removes the evidence of where the image came from, not the fact. Platforms that ask you to disclose AI content still expect you to.
What this doesn't protect against
- SynthID. Removing metadata leaves the watermark in the pixels, and Google's checks in Gemini and Search still read it. We don't offer a way around it, and this site doesn't help with that.
- Classifiers. Platforms and detectors can judge an image from its pixels alone, with or without metadata. See how to check if an image is AI-generated.
- Copies already shared. Cleaning your file doesn't change copies already posted or sent.
- Your Google account. Your chats and generations live in your Gemini activity, separately from the file.
Common questions
Do all Gemini images have C2PA metadata?
Not necessarily. Google announced C2PA for Nano Banana Pro images in the Gemini app, Vertex AI and Google Ads in November 2025, and describes pairing it with SynthID for Nano Banana 2. It hasn't published a list of every product and model that writes it. Check the file itself.
Does a Gemini image include my location?
Google's announcements don't mention location data, and the developer report above doesn't describe GPS fields. A generated image has no camera location. If you edited one of your own photos in Gemini, check the result in a metadata viewer before sharing it, the same way you would any photo.
Can I tell a Gemini image from a ChatGPT image by its metadata?
Usually, if the metadata is still there. The C2PA record names the tool that signed it. What ChatGPT embeds in images shows OpenAI's version. Once the metadata is gone, only pixel-based checks such as SynthID are left.
Why does my viewer show no EXIF for a Gemini PNG?
Probably because of the nonstandard Raw profile type APP1 chunk described above. Many viewers skip or fail on it. ExifTool reads PNG text chunks and is the most reliable way to see it.