AI for GAMES+Direct answer
What is generative game design?
Generative game design uses AI to help people propose, explore, compare, and revise game concepts and content. Systems can generate levels, maps, quests, stories, dialogue structures, visual or 3D assets, rules, mechanics, and broader design alternatives from text, sketches, examples, or partial specifications.
Generative AI for games extends procedural content generation with pretrained knowledge and flexible language or multimodal interfaces. The central challenge is not simply producing an attractive artifact, but producing one that functions inside a game and remains controllable by a designer. A level must be playable and satisfy structural constraints; an asset may need usable geometry, animation, collision, and placement data; a proposed rule must interact coherently with the rest of the system.
Useful tools should expose alternatives, preserve intent across revisions, and support editing instead of forcing designers to accept one opaque output. Evaluation therefore considers validity, diversity, constraint satisfaction, novelty, controllability, editability, and effects on real design work and player experience. Design claims should also remain distinct from implementation claims: describing a mechanic is not the same as building and testing it in an engine.
This topic is surveyed in AI for Games in the Foundation Model Era.
Research scope
What this topic covers
Assets, levels, and worlds
Functional visual, spatial, and structural content that can be placed, played, or revised.
Rules and mechanics
Systems that propose constraints, interactions, objectives, dynamics, and game concepts.
Narrative and co-creation
Mixed-initiative tools for stories, quests, dialogue structures, ideation, and iteration.
Evidence
What should evaluation establish?
Evaluation should test whether outputs satisfy game and designer constraints, remain meaningfully different, and support revision. User studies are strongest when they follow authentic design work and measure correction effort, retained alternatives, and downstream player experience.
- Validity and constraint satisfaction
- Diversity without trivial variation
- Designer control and editability
- Effect on workflow and player experience
Starting points