SynthID detector: how to check for Google's AI watermark
SynthID is the invisible watermark Google puts into images, video and audio made with its AI. Only Google's tools can read it, and wedetect cannot. Here is how to check with Google's tools, what the answer means, and how to check the images SynthID can't speak for, with the free detector below.
How to check an image for SynthID
- Open the Gemini app at gemini.google.com or on your phone, signed in with a Google account.
- Add the file and ask: "Was this image created or edited by Google AI?" The same works for videos under 90 seconds and audio under an hour.
- Read the answer using the guide below. Google's help page notes a daily limit of around ten image checks.
Google also runs a standalone SynthID Detector portal, launched in May 2025 for early testers (journalists, media professionals and researchers) through a waitlist. As of September 2026 the Gemini app is the way most people can check.
What the answer means
- SynthID found: all or part of the file was made or edited with Google's AI.
- SynthID not found: Google's AI did not make or edit it. That is all it tells you. The image can still come from ChatGPT, Midjourney, Stable Diffusion, Flux, Grok or any other generator.
- Unclear: the image had too little detail to carry the watermark, or the AI edit was too small to tell.
When SynthID can't answer, check the pixels
Any image made with another company's tools comes back "not found", so on its own that answer tells you nothing. The detector below doesn't look for any watermark. It looks at the image itself and returns the probability that any AI generator made it. It catches about 85% of images from Midjourney, DALL·E, SDXL and Flux, and it gives an estimate, not a verdict.
Drag & drop, paste, or click to upload
PNG · JPG · WEBP · HEIC — your image is analyzed privately and deleted
This is an estimate, not proof — the detector can be wrong, producing both false positives and false negatives. Don't rely on it as your only source; verify anything important. wedetect.ai is not liable for decisions made from these results. See our Terms.
What SynthID is
SynthID is a watermarking system from Google DeepMind. It hides a signal in the content as it is generated: in the pixels of an image, the frames of a video, the sound of an audio track and the word choices of a text. You cannot see or hear it, and Google designs it to survive common changes such as cropping, filters and compression.
Google adds it across its generative AI consumer products, including the Gemini app. It marks Google's own content only. Other companies use other signals: OpenAI, for example, adds Content Credentials to images made with its tools, which the Content Credentials checker reads.
The other checks worth doing
- Visible marks and metadata. Some generators stamp a mark in a corner or write their name into the file. The metadata checker reads the file, and which AI made this image covers the marks each generator leaves.
- Google's generators specifically. See how to tell if an image is from Gemini and Nano Banana.
- All the watermark and label systems in one place: AI watermarks and Content Credentials explained.
Frequently asked questions
Can wedetect detect SynthID?
No. Only Google's tools can read SynthID. wedetect looks at the image itself and returns the probability that any AI generator made it, which is useful exactly where SynthID has nothing to say.
Is there a free SynthID detector?
The Gemini app checks images, short videos and audio for SynthID for free if you are signed in, within a daily limit. Google's standalone SynthID Detector portal is open only to early testers through a waitlist.
Does "no SynthID" mean the image is real?
No. It only means Google's AI did not make or edit it. The image can still come from ChatGPT, Midjourney, Stable Diffusion, Flux, Grok or any other generator, so check the pixels as well.
Which images carry SynthID?
Images, video and audio made with Google's generative AI consumer products, including the Gemini app, carry it. Text from the Gemini app carries a text version of the watermark.
Can SynthID survive edits?
Google designs it to survive common changes such as cropping, filters and compression. No watermark survives everything, which is another reason a missing watermark is not proof of anything.
Is my image stored when I check it here?
No. The image is deleted right after the check and is never shared.