Gemini’s New Role in Android App Control
Google’s Gemini AI is edging closer to a new level of integration with Android devices, promising to take control of apps on your phone to perform tasks autonomously. Recent discoveries in the Google app beta version 17.4.66 reveal that Gemini will soon offer “screen automation” capabilities, allowing it to place orders online or book rides without requiring constant user input. This development builds on Google’s earlier demonstration of Gemini’s potential at I/O 2025, where the AI was shown navigating apps by scrolling and tapping to complete tasks.
Key Features and User Guidance
The newly uncovered code strings suggest that the feature, internally codenamed “bonobo,” will enable Gemini to interact with popular services like Uber or Lyft to book rides, as well as handle online purchases. However, Google is clear that users must supervise these actions closely. The AI can make mistakes, and users will be responsible for any errors, with the ability to stop Gemini, monitor its progress, or take manual control at any time. This cautious approach highlights the current limitations of AI assistance in automating complex tasks fully.
Privacy considerations are also addressed in the beta code. When Gemini interacts with apps, screenshots may be reviewed by trained human reviewers to improve Google services, but only if users have enabled the “Keep Activity” setting. Google advises against entering sensitive information such as login credentials or payment details during these automated interactions. Furthermore, users are warned not to rely on screen automation for emergencies or tasks involving confidential data, underscoring the need for vigilance when using this technology.
Context and Previous Demonstrations
Gemini’s screen automation feature is part of Google’s broader vision for AI agents on mobile devices, first showcased under the codename Project Astra at I/O 2025. The demonstration illustrated how Gemini could not only interpret on-screen content but also perform actions like tapping buttons and scrolling through apps to complete tasks, effectively acting as a digital assistant with agentic capabilities. Although Project Astra remains under development, these recent beta findings provide tangible evidence that Google is actively working to bring these concepts to everyday Android users.
Alongside screen automation, Google is exploring other Gemini-powered features, including Characters, which lets users create 3D avatars, and enhanced contextual awareness through integration with Maps. These initiatives indicate Google’s ambition to deepen AI’s role in personalising and simplifying mobile interactions.
What to Expect Next
While the beta strings shed light on Gemini’s potential to control apps for ordering and ride-booking, many details remain unclear, such as the full range of supported apps and how seamless the automation will be in practice. The presence of placeholders like a “Purchases” section within Gemini’s interface hints at future in-app order tracking, but these features are yet to be fully realised. As Google continues refining Gemini’s capabilities, users can look forward to more intelligent and hands-free assistance on Android, though caution and supervision will remain important.
For now, the rollout of screen automation marks a significant step in AI-driven phone control, signalling a future where your device could manage routine tasks with minimal input. Observers will be keen to see how Google balances convenience, accuracy, and privacy as Gemini evolves.