Anyone can now check AI video with Google's SynthID Detector

What happened

Google has made SynthID Detector, its tool for spotting AI-generated media, available to anyone. Until now the website was limited to journalists, researchers and other early testers. Since 7 October 2026 it has been open worldwide, in English, at synthid.com.

The bigger change is who it covers. SynthID started as Google's own invisible watermark, built into images, video and audio made with its AI models. The detector now also recognises the watermark when it is used by OpenAI, NVIDIA and Kakao, so one upload can check media from several of the biggest AI image and video tools.

The key details

• What it checks: images, video (including MP4, MOV and WEBM) and audio files. Text is not part of the public site.
• Whose AI: Google, OpenAI, NVIDIA and Kakao. Google says Apple will be added soon.
• Result: a simple answer on whether a SynthID watermark was found. The public version does not highlight which part of a file was edited.
• Access: free, after signing in with a Google, Apple or OpenAI account.
• Limits: roughly 10 checks per person per day, according to Google, partly to stop people using the checker to build watermark-removal tools. Bulk checking by software is not allowed.
• Not a lie detector: Google says plainly that this is not a general AI detector. Media from tools that do not use SynthID will not be flagged, and a clean result does not prove a file is human-made.

The background

Google DeepMind introduced SynthID in 2023 as a watermark that hides inside the pixels or sound of AI output, invisible to people but readable by software, and designed to survive common edits like cropping, compression and filters. It has since been added to Google's image, video, music and text models, and Google has said that well over 10 billion pieces of content carry it.

The detector website was first shown in May 2025 for a small group of testers, and Google later added SynthID checks inside the Gemini app. Opening it to everyone, and reading other companies' watermarks too, comes as AI images and video spread fast across social media and advertising. Platforms, regulators and industry groups are all pushing for clearer labelling of AI content, alongside open standards like C2PA content credentials.

What it means for brands

AI-made video is now something your customers can check in a few seconds. That is good news for honest brands: it makes deepfakes and fake product claims easier to catch, and it rewards being open about using AI rather than hiding it.

For AI ads, the practical step is simple: assume any AI shot can be identified, label AI content where platforms ask for it, and make the creative good enough that nobody minds how it was made. Strong ideas, real product details and a clear brand story matter more than ever. That is how we approach our AI video ads.