What you will build
This guide walks through creating a small Android app with Google AI Studio. The example is a simple community task tracker where a user signs in, adds a task and sees the task again on another device.
Google announced its full-stack Android workflow on November 17, 2026. It uses Kotlin and can connect Cloud Firestore, Firebase Authentication, Firebase AI Logic and Firebase App Check. Google says these features are rolling out, so the exact controls may not yet appear in every account.
Important distinction: Google’s personal Google Play library flow is for installing an app for personal use. It is not the same as a public Play Store launch for customers.
Before you start
You need a Google account, a modern browser and access to Google AI Studio Build at https://aistudio.google.com/apps. You should also have access to the Firebase services that your project needs. Some services may require billing configuration depending on usage and region.
Start with a small app. Do not begin with payments, medical records or sensitive personal data. The aim is to learn the workflow safely.
Step 1: Open the Android workspace
- Sign in to Google AI Studio.
- Open the Build workspace.
- Look for the Android option or Android chip at the top of the workspace.
- Select it if it is available in your account.
- If you do not see it, the feature may not have reached your account yet. Do not assume a missing control means your account is misconfigured.
Step 2: Describe one clear app
A vague prompt produces vague software. Describe the user, the main job, the screens and the data that must be saved.
Try this starter prompt:
Build a simple Kotlin Android task tracker for a neighbourhood volunteer group.
Users should sign in with Google. Each signed-in user can create,
edit and mark their own tasks as complete. Store tasks in Cloud
Firestore so they remain available across devices.
Create three screens: sign in, task list and add or edit task.
Use a clean accessible design with large readable text.
Do not add payments or collect unnecessary personal information.
Explain the data model and Firebase security rules before finalising.
Ask the agent to explain the plan before it makes major changes. Confirm that the proposed data model matches the app you intend to build.
Step 3: Review the backend setup
When the prompt requires persistent data or sign-in, Google says the Build agent can propose and provision Firebase services with your confirmation.
Review each part:
- Cloud Firestore stores tasks so they can sync across sessions and devices.
- Firebase Authentication manages Google Sign-In.
- Security Rules should ensure a user can access only their own records unless the app explicitly supports shared groups.
- Firebase AI Logic is relevant only if the app calls Gemini.
- Firebase App Check helps protect backend resources from unauthorised clients.
Do not approve security rules you do not understand. Ask the agent to explain who can read, create, update and delete each record. Test with two separate accounts.
Step 4: Generate and inspect the Kotlin app
Let the Build agent create the project. Google says the workflow uses the Android CLI and Gradle for project scaffolding and builds.
Ask it to explain the main files, data model, authentication flow and error handling. Check that sign-out works and that the app does not expose another user’s information.
A useful follow-up prompt is:
Review the generated Kotlin project for security and reliability.
List the screens, data collections, authentication flow and permissions.
Identify any hard-coded secrets, unsafe Firestore rules or missing
error states. Do not make changes until you explain the findings.
Step 5: Run the Android emulator
Use the emulator in the Build workspace to test the app. Google describes a faster HD emulator with real-time audio output.
Test the main journey, not just whether the home screen appears:
- Sign in with a test account.
- Create a task.
- Edit and complete the task.
- Close and reopen the app.
- Check that the task remains.
- Sign out and sign in with a second account.
- Confirm the second account cannot see the first user’s private tasks.
If the app uses audio, test it separately and check the device volume and permission behaviour.
Step 6: Improve the interface with screenshots
Google says you can capture a frame from the running emulator and attach it to the coding agent’s prompt.
Capture a screen that needs improvement and give a specific instruction, such as:
Use this screenshot to improve the task list screen.
Increase contrast, make the primary action easier to find,
keep text readable on a small Android phone and preserve
the existing data and navigation behaviour.
Change one area at a time. After every visual change, repeat the functional tests. A screen that looks better can still break navigation or hide an important control.
Step 7: Test on a real phone
The emulator is useful, but it cannot reproduce every device condition. Install the app using the available personal Play library workflow if it appears in your account.
Test on a physical Android phone with a normal network connection and, where relevant, a weak connection. Check screen sizes, keyboard behaviour, sign-in, offline behaviour and data synchronisation.
Step 8: Publish to your personal Google Play library
Google says AI Studio can build and sign an Android App Bundle and publish it to a personal Google Play library using a standard Google account. Follow the publish panel shown in your account and review the app identity and requested permissions.
This personal library process should not be confused with a public Play Store release. A public release has separate developer account, app listing, policy, testing, data safety and review requirements. Check the current Google Play Console requirements before planning a commercial launch.
Step 9: Check costs, privacy and security
Before sharing the app, answer these questions:
- What personal data does it collect?
- Can users delete their data?
- Are database rules restricted by user identity?
- Are API keys and secrets kept out of the APK?
- What happens if Firebase or Gemini is unavailable?
- Have you checked likely Firebase and model usage costs?
- Is there a privacy notice that accurately describes the app?
Never use real customer, patient, financial or student records while experimenting.
Common problems
The Android option is missing. The feature is rolling out. Check again later rather than following instructions for a control you cannot see.
The app forgets data. Check whether the app is using local state or Firestore, and confirm that the user is signed in.
One user can see another user’s records. Stop testing with real data. Review Firestore Security Rules and test with separate accounts.
Gemini calls fail. Check Firebase AI Logic configuration, App Check, project setup and service availability. Do not paste a private key into client code.
The emulator looks different from your phone. Test on the actual device and adjust layouts for different screen sizes.
A sensible first project
For a first attempt, build a simple event registration list, a volunteer task tracker or a personal reading log. Keep the first version to sign-in, one main workflow and a small data model.
The real skill is not writing a long prompt. It is defining a narrow user problem, checking the generated code and proving that the app behaves correctly with real users.
Feature availability and the workflow described here are based on Google’s November 17, 2026 announcement. Controls may vary during rollout.