In This Article
Follow BestForAndroid
What’s the point of an AI detector that only works on some AI? Google just opened SynthID Detector to everyone, and the answer to that question is harder to find than the detector itself.
Google launched SynthID Detector publicly on October 7, 2026. Anyone can now upload an image, video, or audio file at synthid.com to check whether it carries an invisible AI watermark. The tool was previously restricted to journalists, researchers, and media professionals on a waitlist.
Think about the last time a photo made you pause mid-scroll because something felt slightly off, the lighting too clean, a hand with one extra finger. That doubt is what SynthID Detector is built for. Until now, only a limited group of vetted professionals could use it.
How SynthID actually detects AI content
Most AI text detectors try to analyze writing style and guess based on patterns. That track record is shaky, and it’s why they’ve wrongly flagged real students and journalists. SynthID works differently. When Google’s AI generates a picture, video, or audio file, it embeds an invisible mark you can’t see or hear. The detector looks for that mark rather than trying to analyze whether something “looks” like AI.
Google says the marks survive common changes like cropping, filters, and audio compression. The system already runs inside Gemini, Search, and Chrome, handling over a million verification requests each day. The new public website brings that same system into one place anyone can reach.
Which AI tools SynthID can and cannot detect
This is where expectations need to come down a little. The detector works for Google’s AI and a growing list of partners: OpenAI, NVIDIA, and Kakao. Apple support is listed as coming soon. But content generated by any other tool, including open-source models that skip watermarking entirely, won’t show up at all. Google’s own site is direct about it: this is not a general AI detector.
There’s also a daily limit of around 10 checks per user, put in place to make it harder to repeatedly test files while trying to strip the watermark out.
Does a clean result mean the file is real?
No, and Google doesn’t pretend otherwise. A file could come from any AI tool outside the SynthID network. Heavy editing can also damage or weaken an existing mark. And a positive result only tells you AI was involved, not how much. A photo with one AI-removed object and a fully synthetic image register the same way.
Google says to treat the result as a clue, not a verdict, and to cross-check with other sources before making any decisions. That’s honest framing, and it’s the right way to use this.
How to get the most out of SynthID Detector
Upload the best quality version of the file you can find. Every repost, screenshot, and recompression weakens the watermark. The tool supports JPG, PNG, WEBP, GIF, MP4, MOV, MP3, WAV, FLAC, and several other formats.
A tool that catches some fakes is better than no tool at all. But in a world where open-source AI generates millions of images with no watermark system behind them, SynthID Detector covers a smaller piece of the problem than the public launch might suggest. Use it. Just don’t stop there.
