Google Gemini's Mac app includes a screen-sharing feature that transforms the AI assistant into a real-time tutor, coaching users through complex software tasks rather than taking control. In practice, the tool walked a user through editing a screenshot in Photopea—a free Photoshop alternative—by providing step-by-step instructions after viewing the browser window.
The feature, available through the app's Share Window tool, emerged quietly with Gemini's Mac app launch in April. Unlike other desktop AI agents that increasingly automate actions, Gemini focuses on empowering users to learn tasks themselves. Access requires enabling extra privacy permissions under Settings > Privacy & Security > Screen & System Audio Recording.
This educational approach positions Gemini differently from rivals such as Anthropic's Claude or Microsoft's Copilot, which emphasize direct computer control for automation. The tool reflects a broader strategic bet: that users prefer guided learning over handing over autonomy to AI—a philosophy that could resonate in enterprise and education settings.
While promising, the feature's privacy implications may give some users pause. Granting screen and audio recording permissions to an AI assistant tied to Google's cloud raises valid concerns about data handling. The company has not detailed how shared screen data is processed or retained.
Varying sentence structure and focusing on the most newsworthy detail—the live walkthrough anecdote—highlights the practical utility of this under-the-radar capability. The single source, Fast Company, provides firsthand experience but lacks independent verification of user adoption or broader impact.