Mountain View, California, April 19, 2026, 10:31 PDT
Google is updating Gemini with a new feature: the chatbot can now access a subscriber’s Google Photos library to generate custom AI images. The move brings the company’s consumer AI further into users’ private account data, moving beyond just typed prompts.
Personal Intelligence, Google’s opt-in tool for linking Gemini with Google apps, now works with Nano Banana 2, the company’s latest image-generation model. According to The Verge, the system taps Google Photos labels to pick out people—like the user, friends, and family—then spins up an image from just a basic prompt.
Timing is key here: Google wants to leverage personal data to gain an edge in the paid AI battle. Image generation is now a main entry point for consumer AI subscriptions. Instead of simply upgrading its prompt box, Google is tapping into its vast Photos archive to reduce friction for users looking to create personalized images.
Google said it’s pushing out the update over the next few days to U.S. subscribers on Google AI Plus, Pro, and Ultra. The company plans to add the feature to Gemini in Chrome on desktop, and roll it out to more users after that. According to TechCrunch, there’s a “sources” button as well, letting users check how Gemini put together context for any generated image. TechCrunch
Google’s Animish Sivaramakrishnan and David Sharon say the update is about helping users “spend more time creating and less time explaining.” They point to prompts like asking Gemini to sketch a dream house or whip up a picture featuring must-have items for a desert island, with results tailored to preferences pulled from Google apps. Blog
With this update, users skip hunting down a photo, downloading, then uploading it just to create a family-style image. As long as Photos labels are in place, Gemini grabs a suitable image for reference. If it gets something wrong, users can swap in a different reference shot or fix the output directly.
Trust is the sticking point. Google maintains that Gemini doesn’t train on private Google Photos libraries and users can disable app connections. Still, a Google Photos help page notes Gemini features in Photos may tap photos, videos, face-group labels, and Google Account data—processing this content to enhance edits or infer details like age or where top face groups might be.
Initial tests show the tool manages to hold onto a person’s likeness even when the setting changes. Eric Hal Schwartz at TechRadar said his fantasy-style image still looked like him—instead of swapping in “a generic placeholder.” Some results, he added, did make strange assumptions. TechRadar
That lines Google up with OpenAI’s ChatGPT Images, Adobe Firefly, and Meta AI—each pitching their own image generation or text-to-image editing products. Google, though, is leaning on a key distinction: first-party context. If users opt in, the tool can tap into Photos and other linked apps.
Personal Intelligence isn’t exactly a fresh concept. Back in January, Google rolled out the feature, giving Gemini the ability to handle personal queries by tapping into apps like Gmail and Photos. At launch, Google noted that these app connections start off disabled, and users have the option to tweak settings, break the links, or wipe chat history.
Still uncertain: how many people will actually pay to link Photos for image generation—and will average users find the source controls clear enough? If the tool grabs the wrong family face, or drags a private snapshot into a prompt, that convenience argument falls apart quickly.
Google is wagering that personalizing its image model will help Gemini stand apart from generic chatbots and become something tuned to individuals’ lives. Users, though, face a simpler fork: opt for sharper, more tailored images, or keep more of their account context out of the AI’s hands.
