Google’s Gemini AI Gains App Control, But Usage is Strictly Limited

Google is introducing groundbreaking screen automation for its Gemini AI assistant, allowing it to independently navigate third-party apps and complete tasks on behalf of users. This capability, initially available on the Samsung Galaxy S26 series in the US and South Korea, marks a significant step towards the long-promised era of proactive AI assistants. However, this newfound freedom comes with a strict quota system, particularly for free users.

The Era of the Acting AI Assistant Begins

For years, tech companies have envisioned AI assistants capable of seamlessly handling everyday tasks like ordering food or booking rides. Google’s approach with Gemini’s screen automation aims to overcome previous integration challenges. Instead of relying on deep integration with each app, Gemini visually interprets screen elements, enabling it to control supported applications directly.

How Gemini Screen Automation Works

Gemini’s screen automation functions by operating apps within a virtual window on the user’s device. Using voice commands – such as “Order my usual coffee” – Gemini opens the corresponding app and executes the necessary steps. The AI scrolls, taps, and types within the virtual window, mimicking human interaction. Currently, Gemini supports Lyft, Uber, GrubHub, DoorDash, Uber Eats, and Starbucks [9to5Google].

Strict Quotas: A Tiered System

Due to the significant server resources required for screen automation, Google has implemented a tiered usage limit system. As of March 13, 2026, the quotas are as follows:

• Free (Basic): 5 automation requests per day
• Google AI Plus: 12 requests per day (€7.99/month)
• Google AI Pro: 20 requests per day (€19.99/month)
• Google AI Ultra: 120 requests per day (€249.99/month)

This tiered system is separate from the cloud-based “Gemini Agent” feature, which remains exclusive to Ultra subscribers [9to5Google]. The limits underscore the economic realities of providing unlimited, proactive AI assistance.

Gemini’s Expanded Quotas: Thinking and Coding

Google has also refined Gemini’s overall quota system, decoupling limits for “Thinking” (problem-solving) and “Pro” (advanced coding) models. This allows users to utilize one model without depleting the quota of the other. AI Pro users receive 300 Thinking and 100 Pro requests daily, while Ultra users receive 1,500 and 500, respectively [9to5Google].

Safety First: Human Oversight Remains

To address security concerns, Google prioritizes transparency and human control. All AI-driven actions occur within a separate virtual window, allowing users to observe the process in real-time. Before completing any transaction – whether it’s a ride booking or a food order – Gemini pauses, prompting the user to review the details and manually confirm the action. Users can also interrupt or take control of the process at any time.

A Turning Point for Mobile Assistants

Analysts view this update as a potential turning point in the mobile market [The Verge]. Google’s visual approach to screen reading bypasses the integration hurdles that have plagued previous attempts at creating truly helpful AI assistants. This positions Google favorably against competitors like Apple and Samsung, whereas the restrictive quota model reflects the high operational costs involved.

What’s Next for Gemini Automation?

The initial rollout is just the beginning. Google plans to expand the feature to its Pixel 10, Pixel 10 Pro, and Pixel 10 Pro XL devices soon. The list of supported apps is also expected to grow, with food delivery services like Instacart already on the roadmap. Google may relax the limits as it optimizes its server infrastructure. For now, users will need to strategically allocate their daily AI actions.