How to Check AI Video from Veo with the SynthID Detector

Veo, Google's video model, embeds a SynthID watermark in the frames of the videos it generates. Google's official SynthID Detector reads that watermark and, for video, marks the time ranges that carry it. This guide covers which tool to use, how to prepare the file and how to read the result.

Which tool to use for a Veo video

• Google's SynthID Detector at synthid.com. Accepts MP4, MOV and WebM files up to 100 MB and marks the watermarked segments. It covers Google and the partners that adopted SynthID, including NVIDIA's Cosmos video models. It needs a sign-in with a Google, Apple or OpenAI account and has a daily quota.
• The Gemini app. Checks Google AI video under 90 seconds, with about 10 video checks per 24 hours. Upload the clip and ask "Was this made with Google AI?"
• Our metadata checker. Reads Content Credentials, which Veo files can carry, and container tags, inside your browser. It cannot read the watermark itself.

Step 1: get the best copy of the video

The watermark lives in the frames, so quality matters. Download the original file where you can, instead of using a screen recording or a copy re-shared through a messaging app: heavy compression and editing weaken the mark. If the file is over 100 MB, trim it to the part you want to check and export it at the highest quality your editor offers.

Step 2: read the Content Credentials

Drop the file into the checker on our homepage. Nothing is uploaded. Video writers often store metadata at the end of the file, so for large files the checker scans the first 6 MB and the last 4 MB, where the metadata sits. If Content Credentials are present, the report shows who signed them and whether they declare AI generation.

Most social platforms strip this metadata on upload, so an empty report is common and proves nothing. The watermark is part of the frames and can survive where the metadata did not.

Step 3: run the SynthID Detector check

1. Open synthid.com and sign in.
2. Upload the clip.
3. Read the result. For video, the detector marks the segments where it found the watermark.

For a clip under 90 seconds that you think came from Google, the Gemini app is a quicker alternative.

How to read a SynthID Detector video result

• The whole clip is marked. The video was likely generated or edited by a SynthID-enabled model. More than one video model uses SynthID, NVIDIA Cosmos among them, so check the Content Credentials before you say it came from Veo.
• Only part of the clip is marked. A short marked stretch inside a longer recording suggests AI material was mixed in, for example one generated shot cut into otherwise real footage.
• Nothing is marked. No confident signal. The video may be real, made with a model that does not use SynthID, or edited heavily enough to weaken the mark. "Not detected" is not proof of authenticity.

Google notes that in very rare cases SynthID can trigger on content without a watermark. Before you act on a single result, look for more evidence, such as where the clip was first posted.

Common questions

Why does a Veo clip from social media show no Content Credentials?

Most platforms strip metadata when a video is uploaded. The SynthID watermark is in the frames rather than the metadata, so the synthid.com check still applies.

Can the SynthID Detector catch AI video from other tools?

Only from models that use SynthID. Video from tools without it will not register through the watermark, although the metadata check may still name the generator if its tags survived. The models list and the SynthID Detector FAQ show what is covered.

What about still images?

For pictures, read Was this image made with Gemini? and Do ChatGPT images have SynthID?

---

SynthID Detector is an independent site, not affiliated with Google. SynthID is a trademark of Google LLC. Limits are as Google published them in October 2026.