Do Gemini and Nano Banana images carry SynthID, a sparkle or Content Credentials?

Google says they do. Its documentation says images from the Gemini API carry an invisible SynthID watermark, and its Gemini app help page says the app adds SynthID and Content Credentials, plus a visible mark that a setting controls. Our checker does not check SynthID, so a result cannot confirm the watermark.

What does Google say Gemini and Nano Banana images carry?

Google describes three separate signals. They live in different places, they fail in different ways, and a file can lose one and keep another.

Signals Google describes for Gemini and Nano Banana images, and what our checker does with each. Checked on October 30, 2026.
Signal	What Google says	Our checker
SynthID	An invisible watermark. Google says all images generated through the Gemini API include one, and that the Gemini app includes one in media it creates or edits.	Does not check it. Every result says so.
Content Credentials	A signed record. Google Cloud says listed models, including several Nano Banana models, sign their output as Google LLC. The Gemini app help page says Gemini Apps include Content Credentials (C2PA) metadata.	Reads and validates it, if the file still has it.
Visible mark	A mark on the picture itself. Google says visible watermarks appear on newly generated images in the Gemini app, and a setting turns them on or off.	Does not look for it. You look.

Is the invisible watermark SynthID?

Yes, according to Google. The Gemini API image generation guide says "All generated images include a SynthID watermark." The same page names the Nano Banana models, including Nano Banana Pro and Nano Banana 2.

For the Gemini app, Google's watermark settings help page says all media created or edited with Gemini Apps includes an invisible SynthID watermark. That page says its settings apply to Gemini Apps only, not to AI features in other Google products.

Google's SynthID page says the watermark is invisible and is added when the content is created. Our guide What is SynthID? covers what the detector does and does not tell you.

Do Gemini and Nano Banana images carry Content Credentials?

Google says so in two places, and they cover different products.

Google Cloud. A Google Cloud documentation page, last updated 2026-10-28 UTC, lists the models that support Content Credentials. They include gemini-2.5-flash-image, gemini-3-pro-image, gemini-3.1-flash-image, gemini-3.1-flash-lite-image and gemini-nano-banana-2.1. It says output from the listed models is automatically signed by Google LLC, with "Google Media Processing Services" named as the app or device.
The Gemini app. The Gemini app help page says Gemini Apps include Content Credentials (C2PA) metadata to provide information about the content's origin. It gives no model list and no description of the record.

The Google Cloud page does not mention the consumer Gemini app, so we do not assume that its details carry over. The page also says C2PA does not classify media as "real" or "fake" and does not guarantee that the statements inside a record are true. The C2PA explainer adds that provenance data can be removed. What are Content Credentials? explains the signature and the trust lists.

On our side of the check: our copy of the C2PA Trust List includes six anchor certificates whose subject names Google LLC. We hold no file signed by Google, so we have not measured what the checker shows for one.

What about the visible sparkle?

Google's watermark settings help page says visible watermarks appear on newly generated images in the Gemini app. A setting called Media Watermark, under Settings on gemini.google.com, turns them on or off. Google says that setting "only controls the visible watermark" and does not affect SynthID or Content Credentials.

Google also says that in India, South Korea and Vietnam the setting appears only with an AI Ultra subscription; without one, visible watermarks are applied automatically. Our visible AI watermarks page lists the mark and how weak a signal it is: a missing mark shows nothing, and anyone can draw one.

What did we find in real files?

We hold 5 real files from Wikimedia Commons categories filed under Gemini, Nano Banana and Imagen: 2 Gemini, 1 Nano Banana and 2 Imagen. We ran them through the checker on October 30, 2026. Third parties uploaded these files, and we do not know how each one was handled before upload.

None carries a Content Credentials record.
One Gemini PNG carries an unsigned XMP label, a digital source type of trainedAlgorithmicMedia. The checker shows "Metadata declares AI use".
The other four read "No supported provenance signals found".

This does not contradict Google's pages. A file passes through uploads, re-saves and conversions before it reaches us, and we cannot say what happened to these. Our checker does not read SynthID, so a watermark may be present in the four files it reports as having nothing. The full set of 32 files is on the accuracy page.

Where can you check for SynthID?

Google says its built-in SynthID checks are in the Gemini app, Search and Chrome. Its Gemini help page says you upload one file of 100 MB or less and ask whether it was created or altered by Google AI. It gives a quota of about 10 image checks in any 24-hour rolling window. Quotas change, so read the page. It also says that, as of October 30, 2026, Gemini can only recognize content created by Google AI tools.

Google's October 7, 2026 post, updated October 29, says its SynthID Detector is open to everyone. Its May 19, 2026 post describes the checks and says Content Credentials checks were rolling out in the Gemini app that day. Our guide Google's SynthID checks in Gemini, Search and Chrome covers the steps. Each of those routes opens another company's service, may ask you to sign in, and uploads your image to Google, not to us.

What this means when you check an image

Open the checker and choose the file. Two headlines apply to what we hold for these products.

"Metadata declares AI use". The file carries an unsigned label that names a generative source. Anyone can add or remove such a label, so it is a label, not proof.
"No supported provenance signals found". The checker found no signed record, no source label and no tool name. This does not mean the image is real or human-made, and it does not mean SynthID is absent, because we do not check it.

If you believe the file came from Gemini or Nano Banana, run the checker, then ask Gemini to check it. For other tools see which AI image generators add watermarks, and for how the invisible marks differ see invisible AI watermarks compared. The pillar page How to tell if an image is AI puts every signal side by side.

Frequently asked questions

Does Nano Banana add SynthID?: Google's Gemini API guide says all generated images include a SynthID watermark, and it names the Nano Banana models.
Can I turn off the Gemini sparkle?: Google's help page describes a Media Watermark setting in the Gemini app. It says the setting controls only the visible watermark. In India, South Korea and Vietnam it appears only with an AI Ultra subscription.
Does New Watermark check SynthID?: No. Our checker reads signed records, metadata and one open invisible watermark. Each result says that SynthID was not checked.
If my checker result says nothing was found, did Gemini make the image?: We cannot tell either way. Nothing found does not mean the image is real or human-made, and it does not rule out Google's tools.