Google launches Personal Intelligence to turn Gemini into an ecosystem assistant

Google has released a major update to its Gemini AI assistant called "Personal Intelligence." This beta feature allows Gemini to pull context directly from a user's Google ecosystem, including Gmail, Google Photos, Drive, and Search history. The update shifts the AI from a general-purpose chatbot to a proactive assistant that uses personal records to answer specific questions.

Connecting the digital dots

Previously, users had to search across multiple apps to find specific information. Gemini can now synthesize data from diverse sources to solve real-world problems. During a pilot demonstration, a user at a tire shop asked Gemini for his minivan’s tire specs. The AI retrieved the trim details from a Gmail receipt and found the exact tire size by analysing a photo stored in Google Photos. When the user reached the counter, Gemini also pulled his license plate number from an old image, saving him a trip back to the parking lot.

Key capabilities included in the rollout are:

• Visual Data Retrieval: Identifying text or objects within personal photos, such as gym equipment brands or vehicle identifiers.

• Complex Reasoning: Combining information, such as checking travel dates in Gmail, cross-referencing prices on Google Flights, and reviewing budget spreadsheets in Drive.

• Tailored Recommendations: Suggesting local coffee shops or travel itineraries based on past preferences found in Search and YouTube watch history.

Privacy and data controls

Because the update involves sensitive personal information, Google has implemented several security protocols. Personal Intelligence is switched off by default. Users must manually opt in and can select exactly which Google apps Gemini is permitted to access.

Google stated that Gemini does not train its foundation models directly on a user’s private Gmail inbox or photo library. Instead, the system uses "Retrieval-Augmented Generation" (RAG) to fetch information only when relevant to a specific prompt. Users can disconnect apps or delete their personalisation history at any time through the Gemini settings menu. To improve transparency, Gemini now includes an "Answer now" button that displays the specific sources used to generate a response, allowing users to verify the data.

Availability and technical requirements

The feature is currently rolling out to Google AI Pro and AI Ultra subscribers in the United States. It is available across the web, Android, and iOS. While the initial beta supports personal accounts only, Google plans to expand the feature to more countries and eventually to free-tier users after further testing.

Google acknowledged that the system still faces hurdles. The AI may struggle with nuanced relationship changes, such as a divorce, or misinterpret specific interests. For instance, a user with numerous photos of golf courses taken for a child's hobby might receive golf recommendations they do not personally want. To mitigate these "over-personalisation" errors, Google allows users to thumbs-down inaccurate connections and regenerate responses without personalisation.