In this article:

• What exactly was launched

• What you need to get started

• Step-by-step app building

• Moving the project to Android Studio

• Limitations at launch

• Context and evolution of the feature

On May 20, 2026, the Google AI Studio team announced that the service now lets you build native Android apps right in the browser and test them on your own smartphone. The feature was unveiled at the Google I/O 2026 conference.

Google AI Studio ― is Google's web platform for working with Gemini models. Previously, in Build mode, it allowed you to generate web apps from a text description. Now the same approach has been extended to mobile development: the user describes an idea in natural language, and the service writes code in Kotlin using the Jetpack Compose toolkit.

The key difference from previous tools ― is that you don't need to install anything. Android development traditionally required an installed Android Studio environment and knowledge of the SDK. Now the path from prompt to a finished native app takes minutes and happens entirely in the browser.

The approach in which an app is created through a natural-language description of the task is called vibe-coding. Google has already applied similar logic in the Antigravity development environment ― we covered it in a separate guide.

Google Antigravity: a guide to the agentic development environment 25.05.2026 Read

The Incrypted editorial team figured out how the new feature works, what you need to get started and what limitations the tool has at launch.

• Google AI Studio has added a mode for building native Android apps. The code is generated in Kotlin with Jetpack Compose, and no local development environment is required.
• The preview runs in a built-in Android emulator right in the browser. The finished app can be installed on a smartphone via a USB cable using the ADB protocol.
• When a Google Play Developer account is connected, the app is published to the internal testing track of Google Play in one click.
• Hardware sensors are supported: GPS, Bluetooth and NFC.
• At launch, the apps are intended for personal use. Firebase integrations ― are planned.

What exactly was launched 

Building Android apps is available in the Build tab. The user selects the "Build an Android app" option and starts describing the task. The service generates a project in Kotlin using current Jetpack Compose patterns ― this is Google's official toolkit for building mobile interfaces.

The apps support integration with the device's hardware sensors ― GPS, Bluetooth and NFC. Google names several use cases: personal utilities, simple social apps, projects using sensors and AI-based apps.

A preview of the built app opens in a built-in Android emulator. It runs in the browser, so there is no need to install a separate emulator on your computer. To test on a real device, the app is installed on a smartphone via a USB cable using the built-in Android Debug Bridge (ADB). 

What you need to get started 

For basic work, a Google account and a browser are enough. To test the app on a smartphone via ADB, you will need a USB cable and developer mode enabled on the device.

To publish to the internal testing track, you need a Google Play Developer account ― it is connected in the AI Studio settings. Registering such an account ― is a separate paid step on the Google Play Console side, and the user must complete it independently.

Step-by-step app building

Let's walk through the process using the example of a simple app for tracking crypto purchases under the dollar-cost averaging (DCA) strategy. The user enters their trades, and the app calculates the average entry price for each asset.

Step 1. Choosing the mode and describing the app

Open Google AI Studio, go to the Build tab and select the "Build an Android app" option. In the input field, describe the task. 

Prompt

In English:

Copy Build a native Android app for tracking dollar-cost averaging (DCA) crypto purchases. Let the user add a purchase entry with: asset ticker (e.g. BTC, ETH), amount spent in USD, price per coin at purchase, and date. Store all entries locally on the device. For each asset, show the total invested, total coins accumulated, and the average entry price. Add a summary card at the top with the overall portfolio cost. Include a delete option for each entry and the ability to filter by asset.

Translation:

Copy Build a native Android app for tracking crypto purchases under the dollar-cost averaging (DCA) strategy. Let the user add a purchase entry with the fields: asset ticker (for example, BTC, ETH), amount spent in dollars, the coin price at the time of purchase and the date. Store all entries locally on the device. For each asset, show the total invested amount, the total accumulated number of coins and the average entry price. Add a card at the top with the total portfolio value. Provide for deleting each entry and filtering by asset.

Step 2. Choosing the design

Before generating the code, AI Studio offers several visual style options for the app ― for example, Clean Minimalism, Elegant Dark, Professional Polish, Vibrant Palette and Sleek Interface. You can choose the option you like in the "Select this design" section or skip the step with the "Skip" button.

Data: Incrypted.

Step 3. Generation and the first build

The Gemini model creates the project ― in our case it is about ten Kotlin files ― and launches the app in the built-in emulator. At the start, the screen is empty: the portfolio counter shows "Total Invested: $0.00", and the list of purchases is empty. 

Step 4. Fixing errors 

If the message "1 error running the code" appears at the bottom of the panel, click the Fix button. The model finds the cause ― in the example it was an initialization error at startup ― and fixes the code. After that the app launches correctly.

Step 5. Testing

Click the "plus" button in the bottom right corner. The "Add Purchase" window will open with the fields Ticker, Amount USD and Price Per Coin. Fill in the trade details and click Add. Enter several purchases ― the "Total Invested" counter at the top will sum up the invested funds.

Step 6. Refining the feature with a prompt

To make the app group purchases by asset and calculate the average entry price, refine the task with the following prompt.

Prompt

In English:

Copy Group the purchases by ticker and, for each asset, add a summary card above its entries showing: total invested, total coins accumulated, and the average entry price. Calculate the average entry price as total invested divided by total coins for that asset. Display it clearly, for example "Avg entry: $2071.67". Keep the existing per-purchase list below each summary.

Translation:

Copy Group the purchases by ticker and, for each asset, add a summary card at the top showing the totals: total invested amount, total accumulated number of coins and the average entry price. Calculate the average entry price as the total invested amount divided by the total number of coins for that asset. Display it clearly, for example "Avg entry: $2071.67". Keep the list of individual purchases below each summary.

After refinement, each asset gets a separate card with the total amount, number of coins and average entry price, and below it ― a list of specific trades.

After testing in the emulator, the app can be installed on a smartphone via ADB over a USB cable or published to the internal testing track of Google Play ― these options are available from the same interface.

Moving the project to Android Studio 

If the project needs to be brought to full publication, it can be transferred to the desktop Android Studio environment. To do this, AI Studio lets you download the project as a ZIP archive or export it directly to GitHub. Further refinement and preparation for publication on Google Play is then carried out in Android Studio. 

Limitations at launch 

At launch, the apps built in AI Studio are intended for personal use. Firebase integrations ― including Firestore, Firebase Auth and Firebase App Check ― are announced for the future.

Installing the app on a smartphone via ADB requires a USB cable and a connection to a computer, meaning there is no fully wireless scenario at launch.

In addition, to publish to the internal testing track, you need a paid Google Play Developer account, which the user sets up separately. 

Context and evolution of the feature 

The launch of Android app building ― is part of a broader set of Google AI Studio announcements at I/O 2026. Alongside it, Google presented the AI Studio mobile app with open pre-registration, integration with Google Workspace, export of projects to the Antigravity environment and image generation for interfaces via the Nano Banana model.

In parallel, Google is embedding app search into AI: a pop-up "Ask Play" panel will appear in the Play Store, and apps will start showing up in the responses of the Gemini assistant. This changes not only how apps are created, but also how users find them.

Source: Incrypted