Gemini presents Personal Intelligence: Google’s new bet on more contextual and personalized AI

Google announced Personal Intelligence, a new Gemini feature that allows the assistant to offer more relevant and tailored responses by connecting—with explicit consent—to services like Gmail, Google Photos, Search, and YouTube. This initiative marks a key step toward more personalized, contextual AI assistants, but it also reignites the debate about privacy and data control.

Google introduced Personal Intelligence, a feature that aims to transform Gemini into more than just an AI-powered search engine. By connecting to user account information, the assistant can better understand routines, interests, and needs, offering more helpful answers without requiring the user to explain everything from scratch. The goal is clear: to move from generic responses to truly contextual interactions.

The proposal is based on a simple idea: if the assistant can access—always with permission—data that already exists in the Google ecosystem, it can reason more deeply. Instead of simply retrieving specific information, Gemini can now cross-reference data, understand situations, and anticipate needs, similar to how someone who already knows your habits and preferences would.

Unlike previous personalization, where the user had to explicitly say «search Gmail» or «check Photos,» Personal Intelligence introduces a qualitative leap: reasoning across multiple sources. Thanks to the new Gemini 3 models, the assistant can combine text, images, and videos to construct more complete and natural responses, without step-by-step instructions.

Google illustrates this capability with everyday examples. From identifying the correct tire size for a vehicle using emails and saved photos, to retrieving a license plate from an image in Google Photos, or suggesting personalized trips based on previous experiences. Its usefulness for recommending books, TV shows, clothing, and vacations is also highlighted, avoiding generic suggestions and prioritizing genuine affinity with the user.

The technical core of this feature lies in Connected Apps, which allow services like Gmail, Photos, YouTube, and Search to be linked. According to Google, the two key strengths are the retrieval of specific details and the reasoning between complex sources, enabling responses that seem tailored to the user and not just statistically calculated.

Regarding privacy, the company assures that control always remains with the user. The feature is disabled by default, requires opt-in, allows users to choose which apps to connect, and can be disabled at any time. Google states that it does not train its models directly with personal emails or photos, but rather uses them only as a specific reference to answer particular queries.

Even so, Google acknowledges risks such as errors, over-personalization, and misinterpretations of emotional and social context. Therefore, it encourages direct user corrections and feedback to improve the system. Access begins in beta in the United States for Google AI Pro and AI Ultra subscribers, with plans for future expansion to more countries and, eventually, to a free version. The challenge will be balancing usefulness, trust, and exposure in an AI that promises to know you better than ever before.

Source: www.itsitio.com