Games and AI have a long history together. Games were early proving grounds for AI research, and “game AI” has always shaped how enemies and characters behave. Now generative AI is changing how games are made and potentially how they are played — while sparking intense debate among developers, artists and actors.

Games as AI research benchmarks

Games offer clear goals, scores and simulated worlds, making them ideal for testing AI:

  • Atari games were the testbed for DeepMind’s DQN in 2013 — see our DQN explainer.
  • Go: AlphaGo defeated Lee Sedol in 2016.
  • StarCraft II: DeepMind’s AlphaStar reached Grandmaster level on the public ladder, as described in Nature in 2019.
  • Dota 2: OpenAI Five defeated the world champion team in 2019.

These systems advanced reinforcement learning techniques now used well beyond games.

AI in game development

Studios use AI throughout production:

  • Code: coding assistants for gameplay programming, tools and debugging.
  • Art and assets: concept art exploration, textures, 3D model generation and variations.
  • Animation: motion generation and cleanup.
  • Audio: voice prototyping, sound effects and music sketches.
  • Writing: drafting dialogue variations, item descriptions and quest ideas.
  • Testing: automated agents that play builds to find bugs and balance issues.
  • Localization: faster translation of in-game text.

AI inside games

  • Smarter non-player characters (NPCs): language models enable characters that hold open conversations, remember players and react dynamically, rather than following fixed dialogue trees.
  • Procedural content: generating levels, quests and worlds.
  • Personalization: adjusting difficulty and content to individual players.
  • Upscaling and frame generation: graphics technologies use AI to render games at higher resolutions and frame rates.
  • AI-generated game worlds: research systems generate interactive, playable environments frame by frame.

The controversy

Generative AI in games is contested:

  • Jobs: artists, writers, voice actors and testers worry about displacement.
  • Performer consent: voice and motion actors object to their performances being used to train or replicate them without consent and pay. Video game performers represented by the US union SAG-AFTRA struck from 2024 to 2025 largely over AI protections, ending with an agreement that included consent and compensation requirements for digital replicas.
  • Quality and originality: players can react negatively to AI-generated assets perceived as low quality.
  • Disclosure: some storefronts ask developers to disclose generative AI use.

What it means for players

In the near term, expect faster development tools, more dynamic characters in some games, and AI-enhanced graphics. Fully AI-generated games remain experimental. The biggest changes may be invisible — in how games are built and tested.

How studios approach AI policies

Many studios have adopted internal policies covering:

  • which tools are approved, and whether their training data and licensing terms are acceptable;
  • where AI may be used — for example prototyping and internal tools — and where final assets must be human-made or reviewed;
  • consent and contracts for performers whose voices or movements are captured;
  • disclosure to platforms and players;
  • security, so that unreleased game content is not uploaded to public AI services.

Platforms have also set rules. For example, Valve’s Steam store introduced a disclosure requirement in which developers describe their use of AI-generated content, distinguishing content made during development from content generated live during gameplay.

AI and accessibility in games

AI also helps make games more accessible: speech recognition for voice commands, automatic captions and audio descriptions, text-to-speech for menus, and adaptive difficulty that lets more players enjoy a game. These uses tend to be less controversial and can expand who is able to play.

What to watch next

  • Commercial games with language-model-driven characters.
  • Disclosure rules for AI content on game stores.
  • Interactive world models generating playable environments.
  • Labor agreements on AI in game development.

Key takeaways

  • Games have been key AI benchmarks, from Atari to Go, StarCraft II and Dota 2.
  • Studios use AI for code, art, animation, audio, writing, testing and localization.
  • Language models enable NPCs that converse and adapt dynamically.
  • Performers struck in 2024–2025 over AI protections, winning consent and pay requirements.
  • AI-enhanced graphics and development tools are mainstream; fully AI-generated games remain experimental.

Frequently asked questions

Do games already use AI-generated characters?

Some games and mods use language models for dialogue, and many studios experiment with them. Most commercial games still rely on written dialogue.

Will AI replace game developers?

AI changes many tasks, but game design, creative direction, quality and coordination still depend heavily on people.

What is AI upscaling in games?

Technologies that render a game at lower resolution and use neural networks to reconstruct a sharper image, improving performance.