Roblox Build Goes Live: AI-Powered Game Creation From a Text Prompt Hits Public Alpha

For two decades, making a game on Roblox meant learning Lua, spending hours in Roblox Studio on a desktop, and mastering a toolchain that was accessible to dedicated hobbyists but impenetrable to casual creators. That changes today.

Roblox began the public alpha of its “Build” feature on July 28, 2026, initially rolling out to age-verified users in New Zealand before expanding globally in the coming months. Build is a mobile-first AI tool that transforms a plain-language description—“a survival horror game in an abandoned hospital” or “a racing game on a floating track in space”—into a fully playable Roblox game, complete with gameplay mechanics, environments, characters, audio, and visual style.

No coding required. No desktop required.

How Build Actually Works

Users access Build through a dedicated tab in the Roblox mobile app on iPhone and iPad. After typing or speaking a game description, several AI models—Roblox has disclosed using a combination of proprietary and open-source foundation models—generate a complete first draft of the game in real time.

The output is not a template. Build produces a coherent game world with working mechanics specific to the genre described: a platformer has gravity and jump physics; a racing game has vehicle controls and lap timers; a survival game has health mechanics and resource collection. The AI makes genre-specific decisions about what a playable game requires, not just what one looks like.

After generation, creators can test the game immediately inside the app and request modifications through a chat interface: “make the monsters faster,” “add a night time setting,” “give the main character a red hat.” Each instruction triggers a targeted edit to the underlying game rather than a full regeneration—a crucial design choice that makes iterative refinement practical rather than frustrating.

Projects created in Build share the same backend, AI model context, and conversation history as Roblox Studio. A creator can start on mobile, bring the game into Studio for deeper customization, then publish from either environment. The seamless handoff between mobile simplicity and desktop power is what Roblox is betting will make Build meaningfully different from the AI game generators that have emerged from smaller studios over the past two years.

Who Can Use It—and Publish

The public alpha in New Zealand has clear age gates. Users must be nine or older to access Build tools at all. Users 16 and older can publish their creations to a global audience through the standard Roblox review process.

Creations that pass Roblox’s safety review—checking for prohibited content, community guideline violations, and IP conflicts—are made available to the platform’s entire user base, which exceeded 400 million monthly active users as of Roblox’s last earnings call. The potential distribution is immediate and enormous, even for a first-time creator who has never shipped anything before.

Pricing is tiered: Roblox confirmed a free base level for Build, with paid tiers for power users offering higher generation limits, priority compute access, and early access to new features. Specific pricing for paid tiers had not been publicly disclosed as of the New Zealand alpha launch.

The Agents Coming Behind Build

Build is the consumer-facing surface of a broader AI infrastructure Roblox has been building internally. Three additional AI agents are in development and expected to reach creators in coming months:

Playtesting Agent: Systematically plays through generated games looking for bugs—stuck geometry, exploitable edge cases, unbalanced difficulty—before a game goes public. The agent provides a structured bug report in plain language that creators can act on without understanding the underlying code.

Analytics Agent: Answers natural-language questions about how a published game is performing: “Why did players stop playing after level three?” or “Which session length correlates with players who come back?” Rather than requiring creators to parse engagement dashboards, the agent synthesizes behavioral data into actionable insights.

Experiment Agent: Recommends A/B tests to improve specific metrics—retention, session length, monetization—and can execute those experiments automatically on a subset of the game’s player population.

Together, the agents form a complete game development lifecycle: generate, debug, publish, analyze, iterate. Roblox is betting that this toolchain will unlock a category of creator that currently doesn’t exist on the platform—people with game ideas and creative vision but without the technical skills or time investment that game development has historically demanded.

Why This Matters Beyond Roblox

The implications of Build extend well past Roblox’s 400-million-user platform. Roblox is the largest user-generated game ecosystem in the world, and making professional-grade game creation accessible to anyone with a phone changes the supply side of that ecosystem in ways that are genuinely hard to predict.

More broadly, Build represents one of the clearest consumer-facing demonstrations so far of what AI-native developer tools look like when they reach mass market. The past two years have been dominated by AI coding assistants like Cursor, GitHub Copilot, and Devin—tools that make professional developers more productive. Build is something different: a tool designed not to augment developers but to replace the requirement for developer skills entirely, at least within a constrained domain.

The constrained domain matters. Game creation in Roblox happens inside a well-defined engine with established mechanics, physics, and platform constraints. Build’s AI doesn’t need to solve general-purpose software engineering; it needs to be excellent at one specific class of creative task. That narrowing of scope is what makes the current generation of domain-specific AI tools surprisingly capable—and what makes Roblox’s version plausible as a mass-market product rather than a demo.

The Competitive Landscape

Roblox is not the only platform pursuing AI-native game creation. Unity has been expanding its AI toolset for professional developers. Snap launched a limited text-to-AR-experience tool in early 2026. Several independent startups, including Ludo and Scenario, have built AI-assisted asset generation tools for game developers.

But none of those competitors have what Roblox has: a captive, enthusiastic creator community of tens of millions of young users who are already deeply familiar with the platform, already have accounts, and are already intrinsically motivated to make and share games with their friends. The moat is not the AI technology—it’s the ecosystem.

Roblox has also been building toward this moment methodically. The company’s AI research team, which it significantly expanded in 2024 and 2025, has been developing the underlying generative models for over two years. Build is not a rushed product launch; it’s the public-facing conclusion of a sustained R&D investment.

Whether Build materially expands Roblox’s creator base—and whether the games AI generates are actually fun to play—will be the real test. The New Zealand alpha gives Roblox a contained population in which to learn those answers before committing to global rollout.

Global expansion is expected in the next few months, with a full launch anticipated before the end of 2026.