You can now upload an image to Gemini to ask if it’s AI-generated

Google just gave Gemini the ability to rat out AI images by integrating its SynthID detection tools directly into the chatbot. We previously saw the tech giant mess around with this specific verification technology earlier this year when they dropped a standalone detector portal.

That old method was a hassle to use, though, because it required navigating a waitlist or manually uploading files to a specific webpage.

Gemini gains SynthID integration

The update changes things by making the process a simple connected app. You can find a new entry in the extensions list inside the bot. You can also just tag the tool directly after dropping an image into the chat window. It is extremely helpful because modern fakes are getting terrifyingly good and you can rarely tell by looking at the pixels anymore.

Even the recent Gemini 3 update made the AI capabilities stronger so we really need a way to spot the fakes. It is admittedly a bit funny that we are using AI to tell on other AI.

SynthID works by looking for invisible metadata that is super hard to scrub out unlike a visual stamp. The system has been used to watermark text and audio and video for a while now.

ALSO READ: Gemini 3 rolls out across Google products with major intelligence upgrades

The best part is that the tool gives you a nuance rating rather than a binary answer. It will tell you if the whole thing is fake or if just a few sections got tweaked by a computer.

Relying on this completely is a bad idea, though. The glaring problem is that the detection only works on images that were watermarked with SynthID in the first place.

Yes, if a generator does not use the specific watermark system then Gemini stays totally clueless. You will have to wait for Google to sign up more partners before this becomes a foolproof solution for everything you see online.