Google’s Gemini is taking a notable step toward becoming a truly individualized AI assistant with the rollout of its new “Personal Intelligence” feature. The update allows the AI to access and process user data from connected Google apps – including Gmail, Photos, and Calendar – to deliver more contextually relevant and tailored responses. currently in beta and available to subscribers of Google AI Premium and Ultra in the U.S., the opt-in feature marks a pivotal moment in the evolution of AI, raising both opportunities and privacy considerations for users.
Google’s Gemini is evolving from a general-purpose AI tool into a more personalized digital assistant with the introduction of “Personal Intelligence,” a new mode that leverages user data from Google apps like Gmail. This marks a significant step toward artificial intelligence capable of understanding individual user needs and preferences. While Gemini remains accessible in its original form, the new engine can now utilize information about you to deliver more tailored responses.
Personal Intelligence in Gemini is initially rolling out in the U.S., with plans for expansion to other regions. Users must manually enable the feature and select which Google applications can contribute data.
Gemini will draw upon data from across Google’s services, including your photos, YouTube viewing history, web search activity (including shopping, news, Maps, flights, and hotels), and productivity apps. This includes Gmail, Google Calendar, and Google Drive. The AI will learn from this data to provide less generic and more personalized responses. The feature is an “opt-in” experience, requiring users to actively enable it through Settings – Personal Intelligence – Connected apps within Gemini. Google will not automatically activate it due to user privacy concerns. Users must individually select the applications from which Gemini can access information.
How Personal Intelligence Works in Gemini:
Google emphasizes that Gemini will not be “trained” on personal data from Gmail or Google Photos. Instead, it will learn from combinations of your requests and where to find relevant information – without accessing your personal data directly.
What This Means for Users
For example, if you ask Gemini about replacing the tires on your car, it will not only know the make and model but can also analyze your photo gallery to understand the types of trips you frequently take. Based on this, it can recommend suitable tires and provide pricing. It can even recognize your license plate from photos, which could be useful for parking payments. Previously, you would have needed to manually provide this information; now, Gemini will “know” it automatically.
Gemini can also identify your preferred vacation types from photos and emails, factoring those preferences into new travel plans – suggesting destinations like botanical gardens, zoos, or playgrounds, and even recommending board games based on photos of past game nights.
You can access personalized information on the Gemini homepage via the “For you” button.
Gemini can suggest TV documentaries based on your viewing history, recommend books based on your reading habits, identify YouTube channels you might enjoy, and even propose surprising gift ideas. Importantly, all of this data is already securely stored within Google’s ecosystem and is not transmitted externally – or to third parties.
Controlling Personalization
If you prefer a non-personalized response from Gemini, simply press the arrow in the regeneration circle. This allows you to disable personalization for that specific response. You can also engage in temporary chats where personalization is not used. Conversely, if you want a fully personalized experience, the “For you” section on the homepage showcases various personalized prompts.
Technical Limitations and Challenges:
- Unclear temporal boundaries (past, future)
- Poor interpretation of relationship dynamics (friend/sibling, mother/grandmother)
- Failure to understand significant life changes (divorce, car accident)
- Incorrect assumptions (recommending a sequel to a book you returned)
- Ignoring user corrections (you don’t eat meat, yet it recommends a meat feast)
- Personal Intelligence is slower; balancing speed and information depth is crucial
- Complex multimodal combinations and outputs
Because of these challenges, Personal Intelligence is launching in beta to refine its accuracy over time.
The beta launch allows Google to address potential inaccuracies. Gemini might sometimes draw connections between unrelated things – for instance, concluding you enjoy golf based on photos of golf courses, without realizing the photos were taken of your children while they were playing. You can correct Gemini by simply telling it you don’t like golf, or by giving a personalized response a “thumbs down,” signaling a misunderstanding. Gemini will also avoid offering health advice or making recommendations related to your health, but will allow you to discuss these topics with it.
The beta version has some areas for improvement. If the AI misunderstands something, simply explain it.
Personal Intelligence is currently available to Google AI Premium and Google AI Ultra subscribers with personal accounts. This launch represents Google’s first major step toward a truly personal and universal assistant, with plans to expand access to the free version and additional countries in the future. It is accessible on the web, Android, and iOS, across all current Gemini 3 models.
