Google personalizes Gemini

Google is letting Gemini use a person’s Google data to make AI-generated images more tailored, extending its “Personal Intelligence” system into image creation. (blog.google) Google said on April 16 that Nano Banana 2 inside the Gemini app can use a user’s interests and preferences, and can also connect with Google Photos to help create images with less prompting. Google’s help pages say Personal Intelligence can connect data from Google Photos, YouTube, Google Workspace apps such as Gmail and Drive, and Google Search services. (blog.google) (support.google.com) Image generation is software that turns text into pictures, and Google says one of the main problems is that users have to describe exactly what they want. The company says personalized image generation is meant to fill in those gaps by drawing on a user’s own context, including style preferences and photos already stored in Google’s ecosystem. (blog.google) (gemini.google) Google has been building that system in stages. It introduced Gemini personalization tied to Search history in early 2025, expanded Personal Intelligence as a U.S. beta in January 2026, and is now applying the same account-linked approach to images. (blog.google 1) (blog.google 2) (blog.google 3) The rollout also deepens the advantage of companies that already hold years of user data across email, photos, video and search. Gemini’s product page says it connects Gmail, Photos, Search and YouTube, plus chat history preferences, to tailor suggestions to “your world.” (gemini.google) Google says users choose which apps to connect and can disconnect them later. Its help documentation says connected-app data can include emails, files, events, photos, videos, saved Search and YouTube data, location information, and “insights” drawn from that material. (blog.google) (support.google.com) That same help page says connected-app data may be used not only to personalize Gemini responses and carry out tasks, but also to improve Google services, including training generative artificial intelligence models, “for everyone.” Google says those settings depend on which apps a user links and the controls they choose. (support.google.com) Google has also been adding provenance tools as its image models spread across products. The company says Nano Banana 2 outputs use SynthID and C2PA Content Credentials, two systems meant to mark or label AI-generated media. (blog.google) For Google, the next test is whether people will trade more account access for better results. The company is now asking users to let Gemini know them well enough to draw the picture before they finish describing it. (blog.google)