What is Google SynthID Detector?
SynthID Detector is Google's verification portal for checking whether media contains a SynthID digital watermark. SynthID itself is a watermarking technology developed by Google DeepMind for AI-generated content.
Unlike a visible logo placed in the corner of an image, SynthID embeds a signal directly into generated content. The signal is intended to be imperceptible to people while still being detectable by Google's verification technology.
Google first introduced SynthID for images in 2023 and subsequently expanded the technology to video, audio and text. The standalone SynthID Detector was introduced in 2025 for journalists, researchers and media professionals before Google expanded access to everyone globally in English in October 2026.
Google says it has now watermarked more than 180 billion images and videos and the equivalent of 240,000 years of audio with SynthID.
Official overview: Google DeepMind SynthID and Google's October 2026 expansion announcement .
How to use Google SynthID Detector
The workflow is deliberately simple. You provide the media file, Google scans it for the SynthID signal, and the detector reports whether a supported watermark was found.
-
Open the official detector. Go to synthid.com. Google DeepMind links directly to this address as the current SynthID Detector.
-
Choose the media you want to inspect. Use an image, video or audio file that you want to check for a supported SynthID watermark.
-
Upload the file to the detector. The portal scans the submitted media for the watermark signal rather than attempting to judge AI origin only from visible artifacts.
-
Allow the scan to finish. SynthID's detection technology looks for the imperceptible watermark embedded when supported AI-generated media was created or altered.
-
Review the result carefully. If the watermark is detected, treat that as evidence that SynthID-enabled AI technology was involved. If it is not detected, do not automatically conclude that the file is real or human-made.
Google describes the portal workflow as uploading content, scanning for SynthID watermarks and reviewing the resulting detection information. Read Google's original Detector announcement .
What does a SynthID Detector result mean?
The most important part of using SynthID correctly is understanding what the result can and cannot establish.
Watermark detected
The detector found a supported SynthID signal in the submitted content. This is evidence that AI technology participating in the SynthID ecosystem generated or altered at least part of the media.
No watermark detected
This does not prove the media is human-made. The file may have come from an AI system that does not use SynthID, or a detectable signal may not be present.
Unclear or partial result
Watermark evidence can sometimes be stronger in particular regions or segments. Google has designed its tools to identify portions of media where a watermark is more likely to be present.
It is specifically designed to detect SynthID watermark signals. That difference matters because many AI tools may generate media without a SynthID watermark.
How SynthID checks images, videos and audio
| Media | How SynthID works |
|---|---|
| Images | SynthID embeds an imperceptible watermark directly into the image. Google says the signal is designed to remain detectable through common changes such as cropping, filters, colour adjustments and lossy compression. |
| Video | SynthID can watermark video frames or segments so that generated or altered portions can later be checked for the watermark signal. |
| Audio | SynthID places an inaudible watermark into supported generated audio. Google says the signal is designed to withstand common changes including noise, MP3 compression and speed changes. |
These watermarks are deliberately different from ordinary metadata. Metadata can be separated from a file or removed during editing. SynthID attempts to place the verification signal directly inside the media itself.
Technical explanation: Google DeepMind SynthID .
Can SynthID Detector identify AI media from OpenAI and other companies?
Google says the expanded standalone SynthID Detector can check supported media made using AI from Google and participating partners. In its October 2026 announcement, Google specifically named OpenAI, NVIDIA and Kakao, and said Apple support is coming soon.
This should not be interpreted as meaning that every image, video or audio file ever produced by those companies automatically contains SynthID. Detection depends on whether the particular product, model or output used a supported SynthID watermark.
| Provider | Status described by Google |
|---|---|
| Core SynthID ecosystem across supported Google AI products. | |
| OpenAI | Named by Google as a participating partner whose supported SynthID content can be checked. |
| NVIDIA | Named by Google as a participating partner. |
| Kakao | Named by Google as a participating partner. |
| Apple | Google says support is coming soon. |
Current partner statement: Google's October 7, 2026 announcement .
The most important SynthID Detector limitations
SynthID can be a strong verification signal, but it should not be used as a final truth machine for determining whether every file on the internet is AI-generated.
- Unsupported AI tools
- AI-generated media can exist without SynthID. A negative result does not rule out generation by another AI model or service.
- Heavy modification
- Google designs the watermark to survive many common edits, but no watermarking approach should be assumed to survive every possible transformation or deliberate attempt to remove it.
- Small or subtle edits
- Very small edits may not always contain enough detectable watermark evidence. Google's Gemini verification guidance explicitly notes that some minor changes can produce unclear results.
- Origin versus truth
- Detecting AI involvement says something about how the media was produced. It does not by itself tell you whether the information shown in the media is factually true or false.
- Partial generation
- A file can contain a mixture of original and AI-generated material. Detection may therefore apply only to part of an image, audio track or video.
Google itself has described SynthID as an important building block rather than a single complete solution for identifying every form of AI-generated media.
Limitations and robustness: Google DeepMind on SynthID .
SynthID Detector vs checking media inside Gemini
Google now offers more than one way to look for SynthID evidence. The standalone detector and Gemini verification are related, but they should not be treated as identical tools.
| Tool | Best use |
|---|---|
| SynthID Detector | The dedicated verification portal at synthid.com. Google says the expanded version can check supported SynthID-watermarked media from Google and participating partners. |
| Gemini verification | Upload media to Gemini and ask whether it was created or edited by Google AI. Google's current Gemini Help page says Gemini currently recognizes SynthID content created by Google AI tools. |
| Google Search and Chrome | Google has also integrated verification features into Search and Chrome to give users more provenance context while browsing. |
In Gemini, Google says you can upload an image, video or audio file and ask questions such as “Is this AI-generated?” or “Was this created or edited by Google AI?”. Gemini then checks available SynthID and, where supported, Content Credentials information.
Gemini verification instructions: Google Gemini Apps Help .
SynthID is different from Content Credentials
Google also uses Content Credentials, based on the C2PA provenance standard. The two technologies answer related but different questions.
| Technology | Main purpose |
|---|---|
| SynthID | Looks for an imperceptible watermark embedded into supported AI-generated or AI-altered content. |
| Content Credentials | Acts more like a digital provenance record, providing information about the media's origin and editing history when compatible credentials are present. |
Using both can be more informative than relying on one signal alone. A watermark can indicate supported AI involvement, while provenance information can add details about where a file came from and what tools modified it.
What should you do if SynthID finds nothing?
A negative SynthID result should be treated as one piece of evidence, not the end of the investigation. If the authenticity of the media matters, use several independent signals.
-
Check Content Credentials or other provenance data. These records can reveal information about the source, creator, camera or editing history when they are available.
-
Reverse-search the image. Finding an older copy or original publication can provide more useful context than visual inspection alone.
-
Inspect the original source. A reposted screenshot, compressed social-media file or edited copy may contain less provenance information than the original media.
-
Examine metadata where appropriate. Original files may contain camera, date, software or other information that disappears after screenshots or repeated sharing.
-
Check the surrounding claim. Even genuine photographs or recordings can be shared with a false caption, incorrect date or misleading explanation.
Common questions about Google SynthID Detector
Is SynthID Detector free to use?
Google says it has expanded the detector to everyone globally in English. Its October 2026 announcement does not describe the public rollout as a paid verification product.
Can SynthID Detector prove an image is real?
No. A missing watermark only means the detector did not find a supported SynthID signal. The image could still have been generated by an AI system that does not use SynthID.
Can it check videos?
Yes. Google's expanded detector supports image, video and audio files. SynthID for video embeds watermark information into generated video content so supported segments can later be identified.
Can it detect AI voices or music?
It can check supported audio for SynthID watermarks. Google has used SynthID with generated audio products including its Lyria technology and other supported audio-generation experiences.
Can it detect every OpenAI image?
Google names OpenAI as a SynthID partner, but that does not establish that every historical or current OpenAI-generated media file carries SynthID. The particular output must contain a compatible watermark for the detector to find it.
Bottom line
Google SynthID Detector gives ordinary users a direct way to inspect images, videos and audio for invisible AI watermark evidence. The public rollout is especially significant because the standalone detector is no longer limited to media professionals and Google says the ecosystem now includes participating external AI companies.
The safest way to use it is straightforward: upload the media, check the SynthID result, and treat the result as provenance evidence rather than an absolute verdict on whether something is “real.”
A positive watermark detection can be meaningful. A negative detection simply means SynthID was not found. For important verification work, combine SynthID with Content Credentials, original-source research, reverse search, metadata and normal fact-checking.
Official resources: SynthID Detector , Google DeepMind SynthID , and Google's October 2026 announcement .