Free Veo Video Checker

What this check can tell you about a suspected Veo clip

The checker answers one question: does the file say it was made with AI? It reads Content Credentials under the C2PA standard, AI labels that use the IPTC digital source type codes, and AI tool names in the file's metadata. It does not say whether Veo, or any other model, produced the clip, and it does not read SynthID. Plenty of tools now make glossy, cinematic footage. Naming a tool is a claim this site will not make.

Your result comes back as one of four verdicts (AI generated, Likely AI, Likely real, or No AI signs found), plus the evidence behind it and still frames from the clip to look at yourself.

For the "was it Google's tool?" part, there is a better route than guessing, because Google marks its own output. That is the next section. Either way, read the result as one signal among several. "No AI signs found" is not a certificate that a clip is real.

Visible marks and SynthID

Google uses two kinds of marks on Veo video, and they behave very differently.

The visible watermark. When Google expanded Veo 3 in May 2025, its post bringing Veo 3 to more countries said it was adding a visible watermark to all videos, except for videos generated by Ultra members in Flow, its tool for filmmakers. So some Veo clips were made without one from the start. And like any mark that sits in the picture, it can be cropped or covered after the fact.

SynthID. The same post says the SynthID watermark is embedded in all content generated by Google's generative AI models. Google DeepMind's post on watermarking AI generated text and video with SynthID explains that for video, the watermark is embedded directly into the pixels of every frame, imperceptible to the human eye but detectable for identification. DeepMind also says plainly that SynthID is not a silver bullet for identifying AI generated content.

How to check for SynthID yourself. Google's post you can now verify Google AI generated videos in the Gemini app says you can upload a video to the Gemini app and ask whether it was generated using Google AI. Gemini scans for SynthID across the audio and visual tracks and says which segments contain elements made with Google AI. Uploads can be up to 100 MB and 90 seconds long.

The two checks answer different questions, so they complement each other. A SynthID check only speaks to Google's tools. A clip from another generator will not carry SynthID, so "no SynthID detected" does not mean "not AI". The file check here reads AI labels whatever the tool, but will not name one, and many clips carry no label at all. The SynthID explainer covers how the watermark works in more detail.

Tells in short cinematic clips

High end video models are good at the big picture: lighting, camera moves, depth of field. The tells tend to hide in the details a cinematographer would never get wrong.

• Camera moves with no operator. Long, perfectly smooth glides through spaces where no crane or drone could fit, with no shake at all.
• Depth of field that cheats. The subject is pin sharp, the background is creamy blur, and objects drift in and out of focus without any reason.
• Speech and room sound. Generated speech can sound too clean for the space: a crowded street with studio quality dialogue and no echo.
• Small hands doing small jobs. Typing, tying laces, counting coins. Fine finger work is still where many clips slip.
• Background text and logos. Shop signs and posters that read as words from a distance and fall apart when you pause.
• Physics in the second half. Many short clips start convincingly and drift: liquid that pours the wrong way, cloth that clips through a body, a reflection that stops matching.

For more tells with examples, see how to spot AI video. If a clip has a faint moving logo of a different kind in the corner, the Sora video checker covers that case.

Clips with a visible Sora mark are covered on the Sora video checker.