What Google says about Veo
Veo is Google DeepMind's video generation model, available through the Gemini app, Google Flow and developer APIs. Google says videos made with Veo are marked with SynthID, its technology for watermarking and detecting AI-generated content. SynthID is embedded in the content itself rather than stored as metadata, so a metadata viewer cannot see it and a visual impression cannot confirm it.
Verify a video in the Gemini app
Google lets people upload a video to the Gemini app and ask a question such as “Was this generated using Google AI?”. Gemini scans for SynthID in both the audio and visual tracks and reports which segments contain elements generated with Google AI; for example, it may find SynthID in part of the audio but not in the visuals. Google says uploads can be up to 100 MB and 90 seconds long. Use the best copy you have, and review the service's terms before uploading someone else's material.
Source: Google: verify Google AI videos in the Gemini app (opens in a new tab)
Visible marks vary
Videos from some Google apps also carry a small visible Veo mark, but whether one appears depends on the product and plan, and a visible mark can be cropped out or imitated. Treat a visible mark as a clue to confirm with a SynthID check, and treat a missing mark as unknown rather than as a sign of a real recording.
What “not detected” means
A SynthID check answers one question: whether Google's watermark was found in that file. A negative result can mean the video came from another generator, that it was made before SynthID was applied in that product, or that the copy was transformed enough to weaken the signal. It does not show that the footage was filmed with a camera. A positive result shows that Google AI was involved in the marked segments; it does not say whether the caption or the claim around the clip is true.
Context still matters
Look for the earliest upload, the account that posted it and any independent coverage of the event it claims to show. Read the platform's AI label and the creator's description. The free Image Evidence checker reads images only, so it cannot inspect a video; a still frame exported from a video will not carry the video's provenance record. Image Evidence has not measured a Veo-specific detection rate.
How do I know if a video was made with Veo 3?
Upload the original video to the Gemini app and ask whether it was generated with Google AI; Gemini checks the audio and visual tracks for SynthID. Then look for a visible Veo mark and check the source of the clip. Keep files under the stated size and length limits, and note any edits made to the copy you checked.
Can Gemini detect videos made by Sora or other generators?
Not through SynthID. Google's help page says Gemini can currently recognize SynthID only in content created with Google AI tools, even though other companies have started adopting SynthID. Gemini on the web and in the Android app can also summarize Content Credentials, which may show a C2PA record from another provider when the file still carries one. For a Sora clip, look for OpenAI's visible watermark and C2PA record.
Source: Gemini Apps Help: verify AI-generated media (opens in a new tab)
Does a SynthID result mean the whole video is AI?
Not necessarily. Gemini can report SynthID in some segments or in only one track, such as generated audio laid over real footage. Read the segment details, and describe the result as precisely as it was reported rather than turning it into a verdict about the entire video.