SynthID Detector is a free Google tool at synthid.com that checks whether an image, video or audio clip carries SynthID, the invisible watermark Google’s AI tools add to what they create. It’s genuinely useful for spotting Google-made (and some partner-made) AI media, but it can’t tell you whether something made with other AI tools is fake.
Key takeaways
- Google opened SynthID Detector to the public worldwide on October 7, 2026 (English only). It’s free.
- You upload an image, video or audio file and it looks for Google’s invisible SynthID watermark.
- It also reads SynthID in media from partners Google names: OpenAI, NVIDIA and Kakao. Apple is listed as “coming soon”.
- The catch: Google says it is “not a general AI detector”. No watermark found does not mean the file is real.
- Use it as one clue alongside reverse image search, Content Credentials and plain common sense.
What is SynthID Detector?
Google has been stamping an invisible watermark called SynthID into media made by its AI tools since 2023. You can’t see it or hear it, but software can. Google says the mark is designed to survive common changes such as cropping, filters and audio compression.
An early version of the detector portal was shown in 2025, but only journalists, researchers and media professionals on a waitlist could use it. On October 7, 2026, Google DeepMind opened it to everyone. According to figures reported from Google’s announcement, SynthID has now been applied to more than 180 billion images and videos.
Google says the same check already runs behind the scenes in the Gemini app, Search and Chrome, handling more than a million verification requests a day. The new website simply gives everyone a direct way to upload a file and ask.
How to use SynthID Detector, step by step
- Get the best copy you can. Save the original file rather than a screenshot. Reposts, screenshots and recompression can weaken the watermark, and Google asks for the highest-quality version available.
- Go to synthid.com on your phone or computer.
- Sign in. Reports from IT-Connect and Engadget say you need a Google, Apple or OpenAI account. It’s free either way.
- Upload the file. Images, video and audio are supported. Common formats such as JPG, PNG, WEBP, MP4, MOV, MP3 and WAV are reported to work.
- Read the result carefully (see the next section). For video, results are shown on a timeline, with the audio and picture tracks checked separately.
One report says checks are capped at about 10 per person per day to stop people from repeatedly testing ways to strip the watermark. Google’s own pages we reviewed didn’t state a number, so treat that limit as unconfirmed.
How to read a SynthID result
This is where most people will get it wrong, so it’s worth slowing down.
- “SynthID was detected.” The file was made or edited with Google AI (or a partner using SynthID). It may have been edited further after that. It does not tell you how much AI was involved. A small AI touch-up and a fully fake image can return the same answer.
- No watermark found. This only means SynthID wasn’t found. The file could still be AI-made with a tool that doesn’t use SynthID, such as many open-source image generators running on someone’s own computer. The watermark could also have been damaged by heavy editing. Or the file is simply real.
Google’s own FAQ admits that “no signal is completely robust” and suggests gathering several data points before drawing a conclusion. That’s the right mindset: a positive result is strong evidence of AI involvement; a negative result proves very little.
The catch: it only finds SynthID
SynthID Detector is not reading the “style” of a picture or guessing from glitches. It is looking for one specific watermark. If the company behind the AI tool didn’t add SynthID, the detector has nothing to find.
Right now Google lists itself plus OpenAI, NVIDIA and Kakao as SynthID users, with Apple coming later and no date given. That covers a lot of mainstream AI media, but not everything. Scammers using local, open-source models won’t show up. Neither will text: the 2025 preview mentioned text, but the public launch covers images, video and audio only.
So who is it for? It’s handy for small business owners checking whether a “customer photo” or supplier image is AI-made, creators verifying clips before reacting to them, and anyone who wants a quick check on a viral image before sharing it in the family group chat. Who can skip it? Anyone hoping for a one-click “is this fake?” button. That tool doesn’t exist yet.
Other ways to check if an image or video is AI
- Ask Gemini. The Gemini app can check uploaded media for SynthID and, according to reporting, also reads C2PA Content Credentials. Same watermark limits apply.
- Content Credentials. Many cameras, Adobe apps and AI tools attach C2PA “nutrition label” metadata. You can inspect it with the free Verify tool at contentcredentials.org. Metadata can be stripped, though, so missing credentials prove nothing either.
- Reverse image search. Google Lens or a similar tool can show where an image first appeared and whether a trusted outlet published it.
- Old-fashioned checks. Who posted it first? Is a reliable news source reporting the same thing? Is someone pushing you to act fast or send money? That last one matters most, as we covered in our guide to spotting fake CRA and Canada Post texts.
Pattern-based “AI detector” websites also exist, but they make educated guesses and can be wrong in both directions. A watermark check like SynthID is more reliable when it says yes, and silent when it says no.
Privacy: what happens to your upload?
You are uploading a file to Google and signing in with an account, so think before you upload anything private, like photos of your kids, client documents or confidential recordings. We couldn’t find a plain-language statement from Google, in the coverage we reviewed, on how long uploads are kept, so check the terms on synthid.com before using it for sensitive material. For businesses in Canada, uploading someone else’s personal information to a third-party service can raise obligations under privacy law, so stick to files you have the right to share.
Station verdict
Worth bookmarking. Trust the “yes”, not the “no”.
SynthID Detector is free, simple and backed by the biggest names in AI, which makes it one of the better tools ordinary people have had for checking AI media. Just remember it’s a watermark reader, not a lie detector. A positive hit is a strong signal. A clean result means “keep checking”. We haven’t scored it because we haven’t run a structured hands-on test; this guide is based on Google’s statements and published coverage.
FAQ
Is SynthID Detector free?
Yes. It’s free to use at synthid.com. Reports say you need to sign in with a Google, Apple or OpenAI account.
Can SynthID detect ChatGPT images?
Google names OpenAI as a SynthID partner, so OpenAI media carrying the watermark can be detected. Older files or heavily edited copies may not be.
Does it work in Canada?
Google says it’s available globally, in English. We found no Canada-specific restriction.
Can it check AI-written text?
Not in the public version. The October 2026 launch covers images, video and audio only.
More on AI and your rights: read how a Japanese court ruled against AI voice clones, browse our list of free AI tools for small business, or see every review in Tools.
Sources: Search Engine Journal · IT-Connect · BestForAndroid · Google (2025 preview announcement). This article was prepared with AI assistance and reviewed by Media Station. Disclosure.