Claude Dominates AI Game Stats: 528 Games Analyzed

Overview

A comprehensive analysis of 528 AI-generated games reveals that Claude is the dominant force in the current wave of generative interactive experiences. According to data collected as of October 2, 2026, Claude was credited on 238 of these titles, accounting for 45% of the catalog. This significantly outpaces competitors like OpenAI’s GPT (20%) and Google’s Gemini (6%).

The dataset, which consists entirely of playable web-based games, offers a unique window into how developers are integrating artificial intelligence into interactive media. The findings indicate that while AI authorship is widespread, the ecosystem remains heavily skewed toward Claude-based tooling and three.js as the primary rendering engine.

"Claude leads, named on 238 of 528 games. The family comes from the creator or from a named model, and one game can name several."

Tool Ecosystem and Engine Preference

The choice of development environment plays a crucial role in the final output of AI-assisted projects. Three.js is the overwhelming favorite for rendering, powering 58% of the catalog with 306 games. This preference aligns with the broader trend of browser-based interactivity, as every game in the dataset runs directly in a web browser without native installation.

Other engines lag far behind. HTML5 Canvas accounted for 6% of titles, followed by Phaser at 5%, and Godot at 3%. Notably, 122 games did not explicitly state their engine, though the browser-native requirement suggests many may rely on standard web technologies.

In terms of AI development tools, Claude Code is the industry standard, appearing in 38% of projects (202 games). Cursor follows with 20%, and OpenAI’s Codex holds 17%. The data shows a strong tendency toward single-tool workflows, with 81% of creators naming exactly one AI assistant. Pairing tools is rare, with the most common combination being Claude Code and Codex, appearing in only 7 games.

Top AI Tools	Usage Count	Percentage
Claude Code	202	38%
Cursor	106	20%
Codex	90	17%
Claude App	26	5%
Google Antigravity	19	4%

Creator Contribution and Code Ownership

A critical metric in this analysis is the proportion of code generated by AI versus human developers. By the creators' own accounts, AI contributed to the vast majority of the codebase in this catalog. 85% of games reported that AI wrote either all or most of the code. Specifically, 81% stated AI wrote "most" of the code, while 5% claimed AI wrote "all" of it. Only 7% indicated AI contributed partially, and 8% did not specify.

This high level of automation suggests that the barrier to entry for game development has lowered significantly. However, the nature of the games themselves remains constrained. 92% of the 528 games were created for game jams, which shapes the scope and complexity of the projects. These time-boxed events encourage rapid prototyping using AI assistants to generate functional loops rather than polished, long-form experiences.

Despite the automation, mobile accessibility remains a challenge. Only 55% of the games are phone-friendly, meaning they can be played in a mobile browser. This indicates that while AI can generate logic quickly, adapting the user interface and experience for smaller screens still requires deliberate human effort.

Genre Trends and Competitive Performance

Genre analysis reveals that Action is the most popular category, comprising 41% of all games (216 titles). Shooter and Simulation games follow at 20% and 18%, respectively. The prevalence of Action and Arcade styles (16%) suggests that AI is particularly effective at generating fast-paced, loop-driven mechanics. Roguelike elements appear in 8% of games, indicating a growing interest in procedural generation, a task well-suited to LLMs.

When examining performance in the 2026 Vibe Coding Game Jam, distinct patterns emerge among top-ranked entries. Of the 42 top-50 games analyzed, 86% ran on Three.js, compared to 68% of all entries in that jam. Furthermore, 69% of the top 50 featured online multiplayer, a stark contrast to the 45% seen across the entire jam field.

"Online multiplayer was a key differentiator for top-performing games, with 69% of the top 50 entries supporting it versus 45% of all participants."

Tool performance within this jam also provides insight into effectiveness. Cursor yielded the best median placement (#230 out of 945 entries), followed by Claude Code (#288) and Codex (#299). While Claude Code is used more frequently overall, Cursor users appeared to achieve slightly higher relative success in competitive settings.

The data underscores a clear hierarchy in the AI gaming landscape: Claude-powered tools and Three.js engines form the backbone of the ecosystem, while multiplayer functionality and mobile optimization remain key areas for future development.