Gemini "Personalized Intelligence" launched in Taiwan: allowing AI assistants to read your Gmail, photo albums, and YouTube to understand you better

Google AI Assistant Gemini’s “Personalized Intelligence Service” is officially open to users in Taiwan today, connecting core applications like Gmail, Google Photos, YouTube, and more, allowing AI to upgrade from “understanding the world” to “understanding you.”
(Background: Gemini app launches “Notebooks,” capable of syncing with NotebookLM workflows)
(Additional context: Google officially releases “Gemini 3”! Climbing to the top as the world’s smartest AI model—what are the highlights?)

Table of Contents

Toggle

• A single command, crossing information walls of four apps
• Privacy design: off by default, full user control
• Phased rollout: subscription users prioritized

Today (April 15), Google announced that Gemini’s Personal Intelligence Service is officially launched in Taiwan and simultaneously opened in the Canadian market, marking another round of global expansion following its initial US debut earlier this year.

The core of this feature is to enable Gemini to no longer rely solely on a general knowledge base but to safely read your Gmail inbox, Google Photos collection, and YouTube viewing history, providing specific, actionable responses based on your real-life context.

A single command, crossing information walls of four apps

The most intuitive expression of the personalized intelligence service is that it breaks down the previous information silos between different Google applications.

For example, planning a trip to Penghu during the Dragon Boat Festival, users only need to ask Gemini, and the system can automatically integrate hotel and flight confirmation emails from Gmail, tagged photos of attractions in Photos, and travel videos viewed on YouTube, generating a complete itinerary and personalized recommendations in one go, without manually switching between multiple apps.

Google explains that this capability is built on two technological pillars: one is logical reasoning across multiple heterogeneous information sources, and the other is the ability to precisely extract details from specific applications (such as emails or photos).

The combination of these allows Gemini to transcend formats like text, photos, and videos, providing responses that are truly close to personal situations.

Privacy design: off by default, full user control

Allowing AI to access Gmail and private photos naturally raises privacy concerns for most users. Google’s design choice is: the connection features are turned off by default, and enabling or disabling them is entirely up to the user, with the option to revoke access at any time.

Once enabled, Gemini will only access relevant data when performing specific tasks, and this data remains securely stored within Google’s existing systems, not uploaded to external servers. In terms of transparency, Gemini will label information sources when generating responses, allowing users to verify and correct them on the spot. Users can also set personal preferences (e.g., “I prefer a window seat”), which the system will remember and apply.

Additionally, users can activate a “temporary conversation” mode within certain dialogues, enabling completely non-personalized, general interactions.

Regarding the most sensitive data training issues, Google explicitly states that Gemini will not directly use users’ Gmail inboxes or Google Photos to train the model. Google will only use prompts entered by users within Gemini and the model’s generated responses to improve service quality.

Phased rollout: subscription users prioritized

The feature is currently being rolled out in phases to eligible personal Google accounts, with priority given to Google AI Plus, Pro, and Ultra subscribers, and expected to gradually expand to free users within a few weeks.

The activation process is straightforward: after seeing the trial invitation on the Gemini main screen, go to “Settings” → “Personalized Intelligence” → “Linked Applications,” and then select the authorized Google services. Once enabled, it can be used on web, Android, and iOS, supporting all models under Gemini.