I was scrolling through a video last week that looked... off. The lighting was too perfect. The hands moved a little strange. Was it AI-generated? I had no idea. And that's exactly the problem Google's trying to solve.
Google just opened up its SynthID Detector to everyone, everywhere, in English. Before this week, it was mostly a tool for media professionals and researchers. Now anyone with a photo, video, or audio clip they're suspicious about can run it through the checker and get an answer.
Here's how it works. Back in 2023, Google started embedding invisible watermarks into AI-generated content through a system called SynthID. You can't see these watermarks. You can't hear them in an audio file. But they're there, baked into the pixels or the sound waves, and Google's detector can pull them out and tell you if the content came from an AI tool.
And we're not talking about a small experiment here. Pushmeet Kohli, who leads science and strategic initiatives at Google DeepMind, said the company has already watermarked more than 180 billion images and videos. Plus 240,000 years' worth of audio content. Let that sink in for a second. Two hundred forty thousand years of audio. I can't even wrap my head around a number that big.
The detector doesn't just catch Google's own AI output either. It works across content made with tools from OpenAI, NVIDIA, and Kakao. Apple's expected to join the list soon too. That's a meaningful detail — these are usually competitors, not collaborators. But when it comes to figuring out what's real and what's not, apparently everyone agreed it's better to work together.
Why does this matter to you? Think about how much of what we see online now started somewhere other than a camera or a microphone. AI-generated videos of politicians saying things they never said. Fake product photos. Audio clips that sound exactly like someone's voice but aren't. A few years ago this stuff was clunky and obvious. Now it's getting good enough to fool almost anyone.
I've been traveling full-time since 2017, and one thing travel teaches you fast is how much we rely on trust with strangers — trusting directions from a local, trusting a review before booking a guesthouse, trusting a photo of a place before you show up. The internet runs on the same kind of trust. When that trust gets shaky, we all slow down, second-guess things, and life gets a little more complicated.
That's the real value of a tool like this. It's not about catching bad actors necessarily. It's about giving ordinary people a quick, free way to ask "is this real?" and actually get an answer instead of just guessing.
Kohli made the point plainly: as AI tools get better and easier to use, having clear context about what you're looking at helps you decide what to trust. That's not corporate spin. That's just common sense.
This detector isn't Google's only move here either. It joins verification features already built into Search, the Gemini app, and Chrome, which Kohli said now handle more than one million requests a day combined. A million. Every single day. That tells you how fast this whole verification question has become part of normal life online, not some niche concern for tech journalists.
So what do you do with this? Next time something looks a little too perfect, a little too convenient, or just plain weird, you've got a free way to check it. Head to the SynthID Detector, upload the file, and see what comes back.
I'm curious to try it myself on some of the stranger videos I've come across lately. Will it catch everything? Probably not yet. But five years ago, this kind of tool didn't exist at all. Give it time, and watch how fast it keeps improving.
What would you check first?
Source: Punch Newspapers
Catch up on our earlier post: Why ChatGPT Trusts the News More Than Your Website
