On October 7, 2026, Google announced global public access in English to SynthID Detector, a tool that checks images, video, and audio for supported SynthID watermark signals. It looks for a clue about a file’s origin—not whether the file’s claims are true. Google’s announcement and SynthID overview describe the tool’s scope.
A watermark is a signal embedded in media; Google says SynthID marks are imperceptible to people but detectable by its technology. Google introduced SynthID in 2023 and said an earlier version of the detector was made available to media professionals in 2025. The public rollout followed that earlier access.
What SynthID Detector checks
Google says the detector checks images, video, and audio for supported SynthID signals. It names Google AI and systems from partners OpenAI, NVIDIA, and Kakao among the sources it supports. The check depends on a file carrying a signal the detector recognizes, so a result is not a universal classification of whether something was made with AI.
Google also describes ways to check content through its other products. In Gemini, a user can upload an image, video, or audio clip and ask whether Google AI created or altered it. Google Search and Chrome also have verification features, though Google describes them as separate routes rather than the public detector portal.
What a result can tell you
A positive result means the tool detected a supported watermark signal. It does not determine whether the content is accurate or misleading, or establish that every part of a file was generated by AI. A negative result means no supported signal was found; it does not prove human authorship. Media made with a system whose watermark the detector does not support may still be AI-generated. La Gaceta’s report also explains these limits.
| Detector result | What it indicates | What it does not establish |
| Supported signal detected | The file contains a SynthID watermark signal the tool recognizes. | Whether the content is true, or whether every part of the file was generated by AI. |
| No supported signal detected | The tool did not find a SynthID signal it recognizes. | That the file was made by a person or contains no AI-generated material. |
Google’s partnership with OpenAI also does not mean that every ChatGPT file carries a detectable SynthID mark. OpenAI operates a separate tool, Verify, which checks supported OpenAI-generated images and audio for C2PA Content Credentials metadata and SynthID signals. Its stated scope is supported content, not every file produced by OpenAI. OpenAI’s Verify documentation describes those checks.
Google also uses the SynthID name for separate work on AI-designed proteins; SynthID Bio is distinct from the media watermark detector.
The watermarking totals Google reported
In its October 7, 2026 announcement, Google said it had watermarked more than 180 billion images and videos and 240,000 years of audio. Those are Google-reported cumulative totals, not counts of checks performed by the public detector.
