On this page

On October 28, 2026 at 04:00 Asia/Shanghai, @GeminiApp reminded signed-in users that Gemini can check whether an image, video, or audio file was made with AI from Google. Prompts such as “Is this AI generated?” or “Is this made with AI?” make the app look for Google SynthID watermarks. Related reading: How to Choose Gemini Video Mode—Static 1 FPS or Agentic Retrieval.

This is an available in-app capability, not a brand-new launch. Google’s May 19, 2026 blog already said SynthID verification for image, video, and audio had recently been added to the Gemini app, and that the feature had been used 50 million times (Google’s own figure). The October post pushes that in-app path.

What Gemini can check

Signed-in users open Gemini Apps or gemini.google.com, upload one file at a time, and ask whether it was created or edited by AI. Help also lists “Was this image/video/audio created or edited by Google AI?”, “Is this AI-generated?”, and “Is this a real video?”

Gemini offers two paths:

  • SynthID — invisible watermarks for Google’s AI models. Detected means all or part was created or edited by Google AI. Not detected means it was not Google AI—but it could still be other AI or ordinary media. Unclear cases include too little detail or an edit that is too small. For video, Gemini reports which portions show a mark; not every frame in a marked portion was necessarily generated.
  • Content Credentials — a C2PA-style digital passport for origin and edit history (AI and non-AI). Help says it is available on gemini.google.com and the Android Gemini app, with iOS soon.

Google-only SynthID in the app

Help is explicit: other companies have started adopting SynthID, but Gemini can currently only recognize content created by Google AI tools. An in-app “not detected” result is not proof the file is human-made, and it is not a scan of every vendor’s watermark. Related reading: DetectorScope Explains Its AI-Text Calls. Why That Still Isn't Proof of Cheating.

Separately, Google Cloud’s AI Content Detection API on the Gemini Enterprise Agent Platform is described as detecting Google and other popular models. That is an enterprise / API line—not the same as SynthID recognition inside the consumer Gemini app.

Limits and quotas

Help caps each check at one file, ≤100 MB, video under 90 seconds, and audio under 1 hour, and asks for tightly cropped screenshots rather than collages. A rolling window of about 24 hours allows roughly 10 image, 10 video (up to about 5 minutes total), and 10 audio checks (up to about 3 hours total).

The May blog also notes expanding detection into Search and Chrome. For Gemini model and pricing context, see Gemini 4 Argon is out globally — $2/$10 now, $4/$20 after November 30. Related trust-and-safety reading: Sweden's Government Offices Block ChatGPT and Claude Sites on Staff PCs—Not a Nationwide Ban.

Details follow Gemini Apps Help and Google’s posts. For reference only.

Related reading