SynthID is an invisible watermark that Google, OpenAI and a few other companies put in the images, videos and audio their AI makes. Here's how to check for it, what the result tells you, and what to do when there's no watermark to find.
🕵️ Check a Photo for AI — FreeFaux Spy isn't affiliated with Google and can't read SynthID. It checks the image itself, so it works with or without a watermark.
SynthID is a watermark made by Google DeepMind. It's hidden in the content itself, so you can't see it in an image or hear it in audio, and deleting a file's metadata doesn't remove it.
Google adds it to content from its own AI tools, such as Gemini and Veo. Google says it has watermarked more than 180 billion images and videos so far. Other companies use it too: Google names OpenAI, NVIDIA and Kakao as partners whose watermarks its detector can read, and has announced Apple as a partner.
SynthID is different from C2PA Content Credentials. Content Credentials are a signed label stored in the file's metadata, saying which tool made or edited it. That label is easy to lose when an image is screenshotted or re-saved. SynthID is built into the pixels, so it's meant to survive more of that.
Only Google's own tools can read SynthID. There are two ways to use them:
Some other websites call themselves "SynthID checkers". To be sure you're using Google's detector, check that the address is synthid.com.
Watermarks can also be stripped. Google says no signal is completely robust, and researchers presenting at USENIX Security 2026 reported attacks that removed SynthID from images while keeping them looking the same.
That's where an AI image detector comes in. Instead of looking for a watermark, it looks at the image itself for the patterns AI generators leave behind. That works on images from any AI tool, whether or not they were watermarked.
Faux Spy's free checker does this: upload a photo or paste a link and you get a verdict with a confidence score in about two seconds. When you check an image by its web address, it also reads any C2PA Content Credentials the file carries. No detector is perfect, so treat the result as one signal, not a final answer.
For the most complete picture, use both:
Checking a dating or social profile? See how to tell if a person online is real.
Upload it to Google's SynthID Detector at synthid.com, which Google opened to everyone on October 7, 2026. You can also upload it to the Gemini app and ask "Was this created with Google AI?" Both check for the invisible SynthID watermark.
No. SynthID only shows up in content from Google and the companies that have adopted it, such as OpenAI, NVIDIA and Kakao. Images from other AI tools have no SynthID watermark at all, and watermarks can be removed. Google's own detector says it is not a general AI detector. To check those images, use an AI image detector that looks at the pixels.
No. Only Google's tools can read SynthID. Faux Spy checks the image itself for the patterns AI generators leave, so it works on images from any AI tool, with or without a watermark. The two checks work well together.
OpenAI is one of the companies Google names as bringing SynthID to its AI-generated content, and Google's SynthID Detector recognizes OpenAI's watermark. An older ChatGPT image, or one that has been heavily edited, may not carry it.
Sometimes. Google says no signal is completely robust, and researchers presenting at USENIX Security 2026 reported attacks that removed SynthID from images while keeping them looking the same. A missing watermark is never proof that an image is real.
3 free checks a day, no account. Use the web checker on any phone or computer, or right-click images with the extension.
🕵️ Check a Photo — Free Add to Chrome — Free