Google has launched Personal Intelligence, a new beta feature in its Gemini app that leverages data from users' Google apps—like Gmail, Google Photos, YouTube, Search, and more—to deliver highly personalized, context-aware responses, positioning Gemini as a more intelligent personal assistant powered by Google's vast ecosystem.
The feature, rolled out last week in the U.S., is currently available to Google AI Pro and AI Ultra subscribers. It is off by default, opt-in, and allows users to select which apps to connect. Google plans to expand it to the free Gemini app and integrate it into AI Mode in Search in the future.
In a blog post, Josh Woodward, VP of Google Labs, Gemini & AI Studio, shared real-world examples of its capabilities:
When Woodward needed tire specs for his 2019 Honda minivan while at a shop, Gemini not only provided the size but recommended options (daily driving vs. all-weather) tailored to his family's road trips to Oklahoma—pulled from photos in Google Photos.
The feature enables Gemini to retrieve specific details from texts, photos, videos, or other sources and reason across complex, personalized data for tailored answers, such as trip planning that avoids tourist traps based on family interests from emails and photos.
However, Google acknowledges limitations in this early beta stage. The model can exhibit "tunnel vision" or over-personalization, inappropriately fixating on certain inferred preferences (e.g., planning an Australia trip overly focused on coffee shops if the user likes them). It may also struggle with timing, nuance, or relationships—such as assuming someone loves golf from many photos when they are actually accompanying a family member. Users can correct these by telling Gemini (e.g., "I don't love golf"), and it will remember and adjust.
A accompanying research paper details the methodology, current challenges (including mixing timelines, overlooking corrections, and misinterpreting relationships), and ongoing work to improve secure data integration, retrieval, long-context handling, and overall personalization quality. Google emphasizes that this is "just the first step" toward a truly personal, universal assistant, with robust privacy controls: Gemini does not train directly on inboxes or photo libraries, and users retain full control over connections and settings.