Google’s Gemini Now Rummages Through Your Photos to Make AI Images

Google started pushing what it calls “personal intelligence” in Gemini earlier this year. The idea was simple: let the AI know who you are, and it’ll give you more relevant answers. Now they’re taking that concept and pointing it at your photo library.

If you opt in, Gemini’s image generator—Nano Banana 2, which I still think is a ridiculous name for what is genuinely one of the better AI image models out there—will dig through your Google Photos. It uses the faces, objects, and labels already attached to your pictures to build more accurate images without you having to spell everything out.

This isn’t entirely new. You could already feed Gemini images of yourself or your dog and ask it to generate something based on those. But it was a manual process. You had to find the right photos, upload them, and hope the model understood what you wanted. Now the bot just goes and grabs what it needs on its own.

The examples Google gives make sense. Instead of typing out a detailed prompt describing your family’s appearance, you can just say “my family at the beach” and let it pull relevant faces from your library. Same for your dog, your car, that weird lamp you bought at a flea market—anything Google Photos has already tagged.

And honestly, it does work better. More personal data almost always means better AI output. The model has more context, more reference points, and it doesn’t have to guess what “my dog” looks like. It can match the breed, the coloring, even the collar.

But here’s the thing I keep coming back to: this is a lot of access. Google is already sitting on one of the most comprehensive collections of personal data on the planet. Now they’re explicitly tying that to an AI that generates images based on your private moments. The opt-in nature of it is reassuring, but we all know how these things go. Once the feature is there and people get used to it, the friction of opting out only increases over time.

I’m not saying don’t use it. The results are genuinely impressive, and if you’re already comfortable with Google’s ecosystem, this is a natural extension of what they’ve been building. Just be aware of what you’re handing over. Your photos, the labels attached to them, the faces your phone has recognized—all of that becomes fuel for the image generator.

For now, the rollout is happening gradually to Gemini Advanced subscribers. If you’re on the standard tier, you’ll have to wait. And if you’re not sure you want an AI rifling through your vacation photos to make a picture of you riding a dragon, the option to keep it turned off is right there.