Gemini's getting maybe its most useful upgrade yet: spotting AI-made content

Do you have a friend who really prides themselves on being able to spot fake, manipulated, composite images? “I’ve seen a few ‘shops in my time. I can tell from the pixels.” Heck, maybe that person is you. But while spotting a quick clone job on an old internet message board was one thing, modern AI has really upped its game when it comes to producing realistic-looking fakes. Thankfully, we’re getting new tools to spot them, and one such solution is rolling out now.

We’re talking about SynthID, which itself isn’t anything new — Google’s been using it to invisibly watermark its own AI-generated and -edited photos, videos, audio, and even text, for a while now. That’s well and good, but how are you supposed to recognize content like that when you stumble across it?

Earlier this year Google announced its SynthID Detector for exactly that purpose, but that was cumbersome from the get-go. Beyond access being initially limited with a waitlist, you had to manually upload suspect media to a web portal for analysis. So while that worked for identifying AI, it was far from streamlined.

Instead of jumping through those hoops, we’ve now got a much easier solution: SynthID recognition has come to Google Gemini.

We’re finding a new SynthID entry appear among Gemini’s connected apps, and engaging with it allows you to quickly scan an image for those tell-tale marks of AI influence. Gemini will report back just how much of the image appears to have had AI’s thumbs all over it, in case you’re dealing with only partially manipulated content.

Useful as this is, there’s still one big underlying problem: Not everyone’s using SynthID, and if some media was generated by an AI model that doesn’t employ it, Gemini’s not going to be able to spot it. Google’s eager to expand the SynthID ecosystem by signing up additional partners, but for the moment, its reach is still a bit limited.