Status

arXiv preprint (arXiv:2607.00527).

Abstract

Generative AI now lets games produce dialogue, quests, characters, images, and worlds at runtime. This paper defines an AI-native game as one in which runtime generative AI is essential to the core loop — removing it would collapse or fundamentally change play. Applying this test to 53 public games and prototypes, we propose a two-axis taxonomy of player-facing game type and dominant AI mechanic, and observe that current titles concentrate in language-driven designs such as narrative adventures. We argue that the central design problem is turning open-ended AI output into stable gameplay through mechanical invariants, and close with a roadmap covering controllable generation, multi-agent systems, evaluation, safety, and regulation.

Citation

Xu, Z., Meng, F., Xu, K., Verbrugge, C., Lucas, S., & Zhao, J. (2026). AI Native Games: A Survey and Roadmap. arXiv preprint arXiv:2607.00527.

@article{xu2026ainativegames,
  author  = {Zhiyue Xu and Fandi Meng and Kaijie Xu and Clark Verbrugge and Simon Lucas and Jian Zhao},
  title   = {AI Native Games: A Survey and Roadmap},
  journal = {arXiv preprint arXiv:2607.00527},
  year    = {2026}
}