AI Watermark Center
Gemini research

Gemini Watermark

Gemini provenance is not one watermark. Visible marks, SynthID, and Content Credentials are separate systems with different inspection paths.

Published
Last verified
Review
Reviewed against primary sources

Gemini provenance is not one single watermark mechanism. A visible Gemini mark, an invisible SynthID watermark, and C2PA Content Credentials are three different mechanisms. Absence of one is not evidence of absence of the others.

Three separate provenance layers
Gemini or Nano Banana output can carry a visible mark where applicable, an invisible SynthID watermark, and C2PA Content CredentialsA generated output splits into three independent branches: a visible mark only where a product surface applies one, SynthID as an imperceptible watermark, and C2PA Content Credentials as structured metadata.Gemini / Nano Banana outputVisible markwhere applicablelogo / overlaySynthIDinvisiblenot AWC-detectableC2PAContent CredentialsJPEG/PNG/WebP inspectable

Visible mark, SynthID, and Content Credentials can each be present or absent independently. Missing one layer does not prove the others are missing.

Visible watermark

Some Gemini Apps outputs can carry a visible mark — a logo or overlay people can see without tools. Gemini Apps Help documents a setting that turns that visible mark on or off for images, video overlays, and music players created in Gemini Apps. The setting does not apply to other Google products. Documented Gemini Apps behavior is not automatically true of Google AI Studio, the Gemini API, Search, Ads, or other Google products.

Absence of a visible mark does not mean SynthID is absent, and it does not mean Content Credentials are absent.

SynthID

SynthID is Google DeepMind’s imperceptible watermark, described as embedded into supported AI-generated images, video, audio, and Gemini text. Users cannot see it. AI Watermark Center cannot read it. AI Watermark Center does not currently detect SynthID. Signed-in Gemini Apps can check supported image, video, or audio files for Google AI SynthID.

Content Credentials

Content Credentials (C2PA) are structured provenance metadata about origin and history. Gemini Apps documentation says Apps include those credentials for AI-created or edited media. That is a metadata layer, not SynthID. This site can inspect embedded credentials in JPEG, PNG, and WebP. Generic C2PA is not, by itself, a Google authorship verdict.

What users can see

  • A visible mark, only if the surface and current setting apply one.
  • Nothing visible for SynthID.
  • Credential details only in a reader that can parse C2PA, including Gemini’s own Content Credentials inspection where supported.

What AI Watermark Center can inspect

  • Embedded C2PA in JPEG, PNG, and WebP: available.
  • Ordinary EXIF, XMP, IPTC, and GPS when present: available.
  • SynthID: not available.
  • Video and audio watermarks: not available.

Public Google verification

Gemini Apps let signed-in users upload a single supported image, video, or audio file and ask whether Google AI created or edited it. SynthID identification there is Google AI only. DeepMind’s separate SynthID Detector portal is still described as an early-tester waitlist, not general availability.

Official Google verification (opens in a new tab)

Inspect Content Credentials

Current limitations

  • Visible-mark rules are Gemini Apps scoped and can vary by region, account type, and plan.
  • Not every Google image model or historical file carries the same signals.
  • Screenshots typically strip C2PA and can crop a visible overlay.
  • This page is not a Nano Banana model encyclopedia — that belongs on the Nano Banana pillar.

Sources

  1. SynthID — Google DeepMind

    Accessed August 16, 2026.

    SynthID modalities (image, video, audio, text). Gemini Apps can check Google AI media. A separate SynthID Detector portal remains an early-tester waitlist, not a general-availability consumer product, as of this access date.

  2. Verify AI-generated images, videos, and audio — Gemini Apps Help

    Accessed August 16, 2026.

    Gemini user-facing verification for image, video, and audio. SynthID identification is Google AI only; Content Credentials inspection is a separate layer.

  3. Manage watermark settings in Gemini Apps — Gemini Apps Help

    Accessed August 16, 2026.

    Current Gemini Apps visible-watermark toggle is Apps-only. India, South Korea, and Vietnam restrict the toggle to AI Ultra; work/school accounts cannot turn it off. The setting does not affect SynthID or Content Credentials.

  4. Nano Banana Pro: Gemini 3 Pro Image model from Google DeepMind — Google

    Published November 20, 2025. Accessed August 16, 2026.

    Historical 20 Nov 2025 launch post: all Google-tool media embedded with SynthID; Gemini-app image verification. Visible sparkle policy in this post is chronological, not a substitute for current Gemini Apps Help.

  5. Start building with Nano Banana 2 Lite and Gemini Omni Flash — Google

    Published June 30, 2026. Accessed August 16, 2026.

    Defines the current Nano Banana family (legacy, 2 Lite, 2, Pro) and states Nano Banana 2 Lite and Gemini Omni use SynthID. Mentions verification through the Gemini app, Gemini in Chrome, or Search.

  6. Nano Banana image generation — Google AI for Developers

    Accessed August 16, 2026.

    Nano Banana is four Gemini API models. All generated images include a SynthID watermark. Does not, by itself, document C2PA for every model.

  7. Gemini 3.1 Flash Lite Image — Google AI for Developers

    Accessed August 16, 2026.

    Nano Banana 2 Lite (gemini-3.1-flash-lite-image): SynthID always on plus C2PA watermarking. Scope is this API model card, not every Google consumer surface.

In this cluster