Strauss Zelnick, CEO of Take-Two Interactive and parent company of Rockstar Games, has made a blunt argument: AI cannot create hit games like GTA 6, despite growing enthusiasm for generative AI tools in the industry. Speaking publicly on the topic, Zelnick finds it “laughable” to suggest that AI systems like Google’s Project Genie—announced in January 2026 as a tool that generates playable worlds from text prompts—can produce blockbuster titles.
Key Takeaways
- Strauss Zelnick is CEO of Take-Two Interactive, parent company of Rockstar Games.
- GTA 6 was postponed to November 19, 2026, with no generative AI in its development.
- Zelnick argues AI tools are “backward-looking” because they rely on historical datasets.
- GTA’s worlds are “meticulously handmade” rather than procedurally generated.
- Take-Two has hundreds of AI pilots across the company but excludes core creative development.
Why AI Is Fundamentally Limited in Game Creation
Zelnick’s position rests on a straightforward claim: creativity cannot exist within any AI model due to its data-reliant nature. AI systems train on historical datasets, making them inherently backward-looking. When Google announced Project Genie as a potential game-creation tool, stock prices for game companies including Take-Two dropped sharply. Zelnick expressed surprise at this negative market reaction, viewing AI as beneficial rather than threatening—but his skepticism about AI’s creative capacity remains firm. The core distinction he draws is between creating assets and creating hits. AI tools may help generate textures, dialogue stubs, or visual elements that resemble existing major releases, but “creating a hit of that magnitude is a completely different animal and does require human engagement and creativity”.
GTA’s open worlds exemplify this gap. These worlds are not procedurally assembled from templates. They are meticulously handmade, with streets and neighborhoods designed with intention, narrative weight, and player psychology in mind. No dataset of existing games contains the architectural blueprint for that level of bespoke world-building. An AI trained on thousands of game assets might produce something that looks like a city block, but it would lack the deliberate pacing, the hidden shortcuts, the environmental storytelling that makes GTA worlds compelling.
Where AI Tools Actually Fit in Game Development
Zelnick is not dismissing AI outright. Take-Two has “hundreds of pilots and implementations” of generative AI across the company, marking a shift in tone from previous skepticism. AI tools could assist with storyboarding, exploring plot alternatives, or generating dialogue variations—tasks that benefit from rapid ideation rather than original creative vision. The distinction is operational: AI excels at acceleration and variation within human-defined constraints, but falters when asked to define the constraints themselves.
Other studios are experimenting with narrower AI applications. Ubisoft has announced plans to use AI to generate dialogue in games, a bounded task where AI can operate within existing character voices and narrative frameworks. Stardock Entertainment used AI to help players create civilizations in Galactic Civilizations IV: Supernova, positioning the tool as an assistant to human players rather than a replacement for design. These implementations treat AI as a subordinate creative tool, not a primary author.
The Myth of AI Leveling the Playing Field
Project Genie’s announcement sparked speculation that AI might democratize game creation, allowing smaller studios or solo developers to compete with AAA publishers. Zelnick’s argument implicitly rejects this premise. The gap between Take-Two and an indie team is not primarily one of asset generation—it is one of creative vision, iterative refinement, and the accumulated expertise of hundreds of skilled developers working in concert. AI might help an indie developer produce polished character models faster. It will not help them conceive a world as rich as Liberty City or design systems as intricate as GTA’s economy.
This does not mean GTA 6 will be immune to AI’s influence. Take-Two’s hundreds of pilots suggest the company is exploring AI’s potential across marketing, analytics, and backend systems. But the creative core—the design decisions that make GTA 6 a hit rather than a competent open-world sim—remains firmly human.
Does AI have any role in modern game development?
Yes, but a limited one. AI can accelerate asset creation, generate variations on existing designs, and assist with routine tasks like dialogue generation or code optimization. It cannot replace the creative vision, narrative judgment, or design intuition that distinguishes hit games from forgettable ones.
Can Project Genie actually create playable games?
Project Genie can generate playable worlds from text prompts, but Zelnick’s argument is that generating a playable world is not the same as creating a hit game. Technical playability and creative excellence are separate challenges, and AI currently addresses only the former.
Why did Take-Two’s stock drop when Project Genie was announced?
Investors feared that AI-powered game creation tools might disrupt the traditional AAA development model and reduce barriers to entry for competitors. Zelnick’s public statements suggest this concern is overblown—the real competitive moat is human creativity, not asset generation.
The debate over AI in game development will persist as the technology evolves. But Zelnick’s position reflects a hard truth: tools that accelerate production are not the same as tools that generate ideas. GTA 6 arrives November 2026, built by human hands and human imagination. AI may eventually prove more capable than Zelnick believes, but until it does, the burden of proof rests with those claiming otherwise.
Edited by the All Things Geek team.
Source: Tom's Guide
