Google AI Studio: Android Apps from Text Prompts

Google AI Studio now generates native Android apps from text prompts — announced May 19, 2026 at Google I/O 2026 Day 1, with AI Studio as the only mainstream prompt-to-app platform emitting real Kotlin/Jetpack Compose code instead of routing through React Native or a WebView wrapper. This post anchors to Google AI Studio's native Android-app generation, announced yesterday (May 19) at Google I/O 2026 Day 1 and deep-dived in today's Day 2 developer sessions (May 20, 2026).

The stakes are real. Every existing "prompt-to-app" tool — Bolt, v0, Lovable, Replit Agent — reaches for React Native via Expo because they started as web platforms. Native Android Kotlin output changes the performance floor, the hardware-sensor story, and the binary-size math in ways that React Native cannot match. For the first time, a developer with no Android experience can ship a hardware-enabled, camera-capable, GPS-aware app directly to the Google Play Internal Test Track without touching a local SDK.

This guide covers everything shipped in the May 19 release: the Kotlin/Jetpack Compose generation pipeline, the Play Console integration, the three initial app categories, an honest 5-way comparison against Bolt, v0, Lovable, and Replit Agent, the free Cloud Run deployment story, the mobile companion app and Antigravity 2.0 hand-off, Workspace integration, and — critically — the gaps that remain as of the I/O Day 2 developer sessions.

The feature lives inside AI Studio's Build tab. Select "Build an Android app," begin prompting, and AI Studio generates production-quality Kotlin code using Jetpack Compose — the official and recommended toolkit for Android UI development per the Android Developers Blog. No Android Studio install, no SDK Manager, no local development environment.

Two preview mechanisms ship alongside the code generator. An in-browser Android Emulator runs inside AI Studio itself for real-time iteration without leaving the browser. For physical device testing, direct USB device install via Android Debug Bridge (ADB) is wired into the AI Studio UI — connect a phone, click install, the app sideloads in seconds.

The companion announcement on the same day was the Android Studio Migration Assistant— an AI-powered tool that ports apps from iOS, React Native, or web frameworks to native Android, converting storyboards, SVGs, and assets while rewriting UI code in Jetpack Compose. Google's framing: "weeks of manual porting into a streamlined agentic workflow that only takes hours." This is a one-way port (to Android, not from it), and a separate tool from AI Studio's generator.

The Play Console integration is the detail most launch coverage glosses over. The flow: connect a Google Play Developer account inside AI Studio, click publish. AI Studio automatically creates the app record, packages the Android App Bundle, and uploads it to the Internal Test Track in Google Play Developer Console. The app becomes available to install within minutes.

The second piece is the auto-update loop. As the developer continues iterating in AI Studio — refining prompts, adding features, adjusting UI — the app on the connected test device updates automatically without manual re-installs. This closes a loop that previously required a developer to trigger a new build, wait for it to process, and manually push an update.

Two features are on the near-term roadmap but did not ship in the May 19 release: test-track management with tester invitations (to share the internal track with other testers), and direct publish to the production or open-testing tracks. The current pipeline ends at Internal Test Track. For any path to public availability on Google Play, developers must use Android Studio or the Play Console web UI directly.

Firebase integrations — Firestore, Firebase Auth, and Firebase App Check — are also listed as upcoming on the Android Developers Blog. None shipped on May 19. Developers who need auth or a real-time database today must wire it in manually after exporting the project.

The May 19 release supports three distinct app categories — each with a different complexity ceiling and hardware profile. Understanding where your use case falls determines whether AI Studio covers it today or whether you need to export and extend in Antigravity or Android Studio.

No prior public comparison includes AI Studio's Android path — it shipped May 19, 2026. Most existing roundups compare two or three tools and miss the native-vs-React-Native distinction that is the single most important variable for anyone choosing a mobile output path. The matrix below covers five tools across eight dimensions readers actually pick on, with each cell sourced to the vendor's own documentation. See also our earlier 4-way comparison of v0, Lovable, Bolt, and Replit Agent — this post adds Google's native-Android entry to that baseline.

The column that matters most for most teams is column 1 (native mobile output). AI Studio is the only entry generating Kotlin — and that distinction is most meaningful for hardware-sensor-heavy apps (camera, GPS, NFC) where React Native's JavaScript bridge introduces latency, and for teams that need platform-native UX fidelity or smaller binary sizes. For pure web development or cross-platform React Native, Bolt, v0, Lovable, and Replit remain valid choices within their respective strengths.

One nuance worth calling out: the broader Q2 2026 20-platform agentic-coding matrix positions these tools against a wider field including Cursor, Windsurf, and GitHub Copilot Workspace. The 5-way matrix above is scoped specifically to the "I want AI to generate my app from a prompt" use case — a narrower decision than which IDE or coding assistant to adopt.

Most launch coverage mentions "two free deploys" and stops. The deeper story is what carries you past those two. Every developer gets two Cloud Run deployments at no cost, no credit card required — a one-time onboarding bonus. After those two deploys, developers with billing enabled transition to Cloud Run's always-free monthly allowance, which carries a toy or early-stage app significantly further.

The two-deploy freebie and the always-free tier are distinct mechanisms with different scopes: the freebie is a one-time developer bonus, the always-free tier is a monthly recurring allowance. Existing Cloud customers can also connect a billed Cloud project to AI Studio for production deploys without the two-app restriction.

Two practical notes from the Cloud Run pricing page. First, the always-free tier applies only in three US regions: us-central1, us-east1, and us-west1. Developers deploying to other regions will not receive the free allowance and should factor standard Cloud Run pricing into their cost model. Second, the two-deploy bonus is per developer (not per project) and is a one-time benefit, not a monthly renewal.

For any team considering how this compares to Vercel's Hobby tier or Replit's free plan, the Cloud Run model is different in character: no always-on dyno equivalent, but a generous monthly request allowance that covers a light-traffic app or a developer testing environment without surprise charges. Our web development and app delivery engagements regularly help teams evaluate these hosting cost structures before committing to a deployment architecture.

Alongside the Android-app generation launch, Google opened pre-registration for the AI Studio mobile app on Google Play on May 19, 2026 at ai.studio/mobile. The app is Android-first; iOS availability follows the Gemini app's historical cadence (typically 1-4 months after Android). The mobile app brings the full Build mode experience to phones — iterate code, preview builds, remix projects from a mobile gallery, and share live deployments. Workspace syncs automatically between mobile and web.

The more significant announcement for developers who outgrow the browser UI is the Antigravity 2.0 hand-off. AI Studio offers three export paths: ZIP file download, direct GitHub export, or one-click hand-off to Google Antigravity . The Antigravity path is unique: it preserves the AI Studio agent state — conversation history, project files, and secrets — so a developer can continue in Antigravity 2.0's desktop app with the same conversation thread and project context intact. ZIP and GitHub exports do not preserve conversation history.

This positions AI Studio and Antigravity as two surfaces in a single funnel: AI Studio for low-friction browser-based generation and quick iteration, Antigravity for teams that need full IDE power, terminal access, and longer-horizon agentic workflows. The lineage of this relationship traces back to Antigravity 1.0 on Gemini 3 Pro — the first time Google shipped an agentic desktop IDE on its own Gemini backend.

One clarification: no SWE-bench or equivalent benchmark numbers have been published for Antigravity 2.0 as of May 20, 2026. The only published Antigravity benchmark is the 1.0 SWE-bench verified score (76.2%). Do not conflate the hand-off capability with performance claims about Antigravity 2.0's code-generation quality.

The Workspace integration is the feature that most clearly separates AI Studio's positioning from every other prompt-to-app tool. Apps built in AI Studio can read and write directly against Google Workspace — Sheets-backed dashboards, Drive-organizing tools, Docs/Gmail-integrated workflows — without separate authentication plumbing. Google handles the OAuth scopes, API connections, and credential management inside the AI Studio surface.

For teams that already operate on Google Workspace — the majority of enterprise and mid-market organizations — this means a developer can build an internal Android app that reads from a Sheets spreadsheet, writes outputs to Drive, and sends notifications via Gmail, all without wiring up separate API credentials or managing OAuth flows manually. The integration is Workspace-level, not limited to any single product.

This is a genuine differentiator. Bolt, v0, Lovable, and Replit Agent can call Google APIs, but they require the developer to manage service accounts, OAuth credentials, and API key handling externally. AI Studio abstracts that for the Workspace layer — a meaningful reduction in integration friction for the internal-tooling use case.

The practical limit: Workspace integration covers data I/O (read from Sheets, write to Drive, trigger Gmail). It does not cover Workspace Admin APIs, Google Meet, or Google Chat as of the May 19 release. Firebase (the logical next layer for Firestore, Auth, and real-time data) remains on the upcoming roadmap.

For agencies managing client data pipelines in Google ecosystem stacks, this opens a class of internal-tool apps that previously required a custom backend. Teams using Gemini 3.5 Flash via the API for data enrichment workflows may find the AI Studio → Sheets integration a faster path to an internal dashboard than a full-stack web app.

Transparent gap-disclosure is part of how we evaluate developer tools. Most launch coverage skips this section; we include it to anchor reader trust and prevent teams from building workflows against capabilities that are not yet shipped. Every item below is sourced from the Android Developers Blog's own upcoming-features list or is a gap observable from the feature announcement.

The gap-disclosure picture is consistent with a Day 1 developer preview: the core generation pipeline, device install, Internal Test Track integration, and Cloud Run deploy are all live. The layers that typically follow — broader store distribution, auth, databases, tester management — are explicitly on the roadmap but not yet shipped. Teams planning a production Android app workflow around AI Studio should design for the current capabilities and build in a hand-off path to Android Studio for anything the May 19 release does not cover.

For the broader landscape of where AI coding agents stand across platforms and capability tiers, see the Claude Code vs Cursor vs Codex vs Replit agentic-coding comparison — the context for how AI Studio's Android path fits into the wider Q2 2026 field.