Google can now spot its own AI videos

Google just made it easier to spot deepfakes, right from your phone.

You can now upload a video directly into the Gemini mobile app and ask whether it was created or edited using Google AI.

Scanning frames and audio for SynthID

The feature works by scanning both the visual and audio tracks for SynthID, Google's imperceptible digital watermark.

Instead of a generic yes or no, Gemini uses its own reasoning to point out specific timestamps. For instance, it can tell you if SynthID was detected in the audio between 10 and 20 seconds while the visuals were untouched.

There are a few limits to keep in mind. Videos can be up to 90 seconds long and under 100 MB. The feature is now rolling out globally in all languages and countries where Gemini operates, alongside image verification.

Why it matters

Putting detection tools directly inside a consumer chatbot lowers the friction for everyday fact-checking.

Instead of relying on niche forensic websites or developer dashboards, regular users can debunk suspicious clips in seconds. It is a smart move that makes provenance tools genuinely accessible.

The catch, of course, is that Gemini only checks for Google's own SynthID watermarks. It will not flag synthetic media generated by OpenAI, Runway, or other competing models, but it sets a practical precedent for platform accountability.