Google’s Gemini app now lets you check a video for Google’s SynthID watermark and see which parts of the clip are flagged.
📌 Key Takeaways
- Gemini can detect Google’s SynthID watermark in videos made or edited with Google AI.
- Detection checks visuals and audio, and can return specific time ranges when found.
- A step-by-step guide below shows how to run a quick verification in the Gemini app.
- Files are limited to 100 MB, and videos should be under 90 seconds for best results.
- “No watermark” means no Google watermark detected, not “definitely not AI.”
What Gemini’s New Video Verification Does
Gemini can scan an uploaded video to see whether it contains Google’s SynthID watermark, which is designed to persist even after common edits.
If detection succeeds, Gemini reports that the watermark is present and can indicate where in the clip it appears, including watermark signals found in audio.
“Gemini will scan for the imperceptible SynthID watermark across both the audio and visual tracks.” — Google
What SynthID Proves And What It Does Not
SynthID is a provenance signal, not a general-purpose deepfake detector. It can be confirmed that a watermark from Google’s AI tooling is present, which is strong evidence that Google AI was used in creation or editing.
A “not detected” result does not rule out AI. It only means Gemini did not find Google’s watermark, and the video could still be AI-generated using other systems, or too heavily edited to retain a detectable signal.
How To Verify A Video In The Gemini App
Use this when you want a quick yes/no provenance check for Google AI involvement, plus any flagged segments.
- Open the Gemini app and sign in.
- Tap Add files, then select a video (from Files, Photos, or Drive).
- Ask: “Was this video created or edited by Google AI?” (or type
@synthid). - Review the result and note any flagged time ranges (audio or visual).
Limits And How To Read The Result
Google’s help documentation lists a 100 MB file limit, and says videos should be under 90 seconds (with additional daily verification limits).
If no watermark is detected, treat it as “no Google watermark found,” not a clean bill of authenticity. That’s especially important for reposted clips, re-encodes, and heavily edited videos.
“If a SynthID watermark isn’t detected, it means the image or video wasn’t created or edited by Google AI, but it could have been created by other AI systems.” — Google
Conclusion
This update gives Gemini a practical, repeatable way to verify whether a video includes Google’s AI watermark and where it appears in the clip.
It’s not a universal AI detector, but it is a useful provenance check you can pair with context, sourcing, and other verification methods when a clip’s origin matters.
📈 Latest AI News
19th December 2025
- OpenAI Launches ChatGPT App Store: Browse, Connect & Use Apps
- OpenAI Releases GPT-5.2-Codex: Most Advanced Agentic Coding Model
- China’s ‘Manhattan Project’ for AI Chips Shocks Global Tech Leadership
- Inside Hollywood’s Creators Coalition on AI
- Gemini 3 Flash Rolls Out Globally as Default Model in Search & App
For the recent AI News, visit our site.
If you liked this article, be sure to follow us on X/Twitter and also LinkedIn for more exclusive content.
