Short answer
A Google AI Studio app is a web app that calls the Gemini API, so the fastest route to Google Play is to host it properly, move the API key behind a server, and wrap the hosted URL in a native Android shell. Then you complete the Play Console declarations, run closed testing and submit.
On this page
- What you actually have after building in Google AI Studio
- Step zero: fix the Gemini API key before anything else
- Step one: host it on a real domain
- Step two: choose your Android route
- Step three: give the Android app real app behaviour
- Step four: the Play Console work specific to AI apps
- Step five: test, submit, iterate
- Timeline you can plan with
- FAQ
- Tools from the jocapps ecosystem
What you actually have after building in Google AI Studio
Google AI Studio lets you describe an app and get a working web application, usually a React and TypeScript front end that talks to the Gemini API. You can preview it inside AI Studio, share it, export the code or deploy it to Google Cloud. What you do not get is an Android project, a signed app bundle or a Play Console listing.
That matters because Google Play only accepts Android App Bundles (AAB). So the question is not "how do I upload my AI Studio app" but "how do I turn a Gemini-powered web app into an Android app that Google is happy to distribute". There are three realistic answers, and the right one depends on how much native behaviour your users need.
Step zero: fix the Gemini API key before anything else
Many AI Studio prototypes call Gemini directly from the browser. In the preview this is convenient. In a public Android app it is a liability, because anything shipped to the client can be extracted. Someone pulls your key out of the app and you pay for their traffic.
Before you think about Google Play:
- Move every Gemini call into a small backend, for example a Cloud Run service, a Firebase Cloud Function or any server you already run.
- Have the front end call your endpoint, not Gemini directly.
- Add per-user rate limits and, ideally, sign-in, so one abusive client cannot burn your quota.
- Restrict the key in Google Cloud to the services it needs and rotate the one used in the prototype.
This also gives you one place to log requests for debugging; if the app handles personal data, decide now what you log.
Step one: host it on a real domain
Export the code (to a zip or a GitHub repository) and deploy it somewhere stable. Deploying to Cloud Run from AI Studio is a reasonable start; Firebase Hosting, Vercel or Netlify also work for the static front end. Put it on your own HTTPS domain. Your Android app will point at this URL, and Google Play requires a privacy policy URL that should live on the same domain.
Check three things on a real Android phone in Chrome before going further: the layout at small widths, how long the first Gemini response takes on mobile data, and what happens when the network drops mid-request. These are the moments reviewers and early users notice.
Step two: choose your Android route
| Route | What it means | Good fit | Effort |
|---|---|---|---|
| Native shell around the hosted app | An Android app that loads your URL in a WebView and adds native features | Most AI Studio apps: chat tools, generators, internal assistants | Days |
| Trusted Web Activity | Package a PWA as an Android app through Chrome | Apps that are already solid PWAs | Days, plus PWA work |
| Native rebuild | Rewrite in Kotlin, Flutter or React Native with a coding agent | Apps needing deep device integration or heavy offline use | Weeks |
For most Gemini-based tools the shell route is the pragmatic one. Your logic and prompts already live on the web, and you keep a single codebase. WebViewGold is built for exactly this: you take its Android app template, insert your hosted URL, configure options such as push notifications, offline handling and native APIs, and build an AAB in Android Studio without writing app code. Because updates to your web app go live without a new store release, you can keep iterating in AI Studio or your own repository.
If you would rather not touch Android Studio at all, WeSetupYourWebViewApp does the WebViewGold setup for you and hands over the full source code.
Step three: give the Android app real app behaviour
Google Play is generally more tolerant of web-based apps than Apple, but its policies on spam and minimum functionality still reject apps that are just a bookmark. Shells that add genuine mobile value fare better. For an AI Studio app, sensible additions are:
- Push notifications when a long generation job finishes or a shared result is ready.
- Native file and camera access if your app analyses images or documents with Gemini's multimodal input.
- A proper offline screen instead of Chrome's error page, explaining that the AI features need a connection.
- Share intent support so users can send text or images from other apps into yours.
- Pull-to-refresh and back-button handling that behave as Android users expect.
WebViewGold covers these through configuration rather than custom code, which keeps the AI Studio project itself untouched.
Step four: the Play Console work specific to AI apps
Create the app in Play Console and work through the policy declarations carefully. For a Gemini-powered app, pay particular attention to:
- Data safety form. Declare what you collect and that prompts are sent to a third party for processing. Be consistent with your privacy policy.
- Generative AI content. Google Play has policies for apps that generate content with AI. Expect to need in-app ways for users to report or flag offensive output, and make sure your system prompt and Gemini safety settings prevent clearly prohibited content.
- Content rating questionnaire. Answer honestly for what the model could produce, not only for what you intend.
- Target API level. Google Play requires new apps and updates to target a recent Android API level. A maintained template saves you from chasing this yourself.
- Test account. If sign-in is required, provide working credentials in the App access section.
New personal developer accounts must also run a closed test with a minimum number of testers for a set period before production access. Organisation accounts are treated differently, which is one reason to register as a company; see our guide on developer accounts for companies.
Step five: test, submit, iterate
Upload the AAB to the internal testing track first, install it from Play on a few devices, and run through the core flow with real Gemini calls. Then promote to closed testing, collect feedback and move to production. Our article on TestFlight and Play Console internal testing covers the tracks in detail.
If you get a policy rejection, read the exact policy cited; the fix is usually a declaration, a missing reporting mechanism or an inaccessible test login. appsubmitter.io handles store listings, submission and rejection responses inside your own developer account if you prefer not to do this yourself. For common failure patterns see Google Play rejection reasons for web apps.
Timeline you can plan with
- Day 1: move Gemini calls server-side, deploy to your domain, write the privacy policy.
- Days 2 to 3: build the Android shell with WebViewGold, add push and file access, create the Play listing.
- Days 3 to 17: closed testing (longer for new personal accounts), fix feedback.
- Then: production review, usually days rather than weeks.
If you also want iOS, the hosting and backend work carries over; the Apple side is stricter, as described in publishing a Google AI Studio app to the App Store.
FAQ
Can I upload a Google AI Studio app directly to Google Play?
No. AI Studio produces a web app, and Google Play only accepts Android App Bundles. You need to host the web app and package it as an Android app, either as a native shell, a Trusted Web Activity or a native rebuild.
Is it safe to ship my Gemini API key inside the Android app?
No. Any key in client code can be extracted. Route Gemini requests through your own backend, add rate limits and rotate the key you used during prototyping.
Does Google Play accept WebView apps?
Yes, as long as the app offers real functionality, works reliably on mobile and meets the usual policies. Adding native features such as notifications, file access and proper offline handling makes approval far more predictable.
Do I need to resubmit every time I change the app in AI Studio?
Not with a shell approach. Web changes go live when you redeploy the hosted app. You only resubmit when you change native features, permissions or the target API level.