With the rapid development of AI, more and more game projects are attempting to incorporate AI or use it as a selling point. At this year’s Game Developers Conference, I encountered countless AI-related lectures, demos, and development tools. This led me to a question: what role should AI actually play in game design?

The first AI-related game I tried at GDC left me deeply disappointed. Before I started to play the demo, I was told that AI played a significant role in this game, but I did not fully understand what that meant. At the beginning, I answered three multiple-choice questions, each with three options. The questions included topics such as my favorite fruit and my level of gaming experience. Within seconds, the AI generated a character and adjusted the difficulty accordingly.

The game itself involved fighting robots, accompanied by an AI assistant. I could actively request help from the AI assistant through a button, or the AI would automatically detect when I needed assistance. It could heal me, lower the difficulty, or even eliminate enemies on my behalf. Throughout the experience, I felt that the player’s role was reduced to simply moving toward enemies, while most of the actual gameplay could be handled entirely by the AI. This excessive intervention significantly diminished the gameplay experience. It made me question: what makes a game a game? When players are no longer required to make decisions or engage meaningfully with the system, can it still be considered a game?

You vs Zombies is another game I played at GDC, it was developed by 10six Games, and offered a very different experience. In this game, players must dodge waves of zombies while attacking them. Here, AI is primarily responsible for character creation. Players can input any prompt, and the AI generates a character’s appearance and attributes based on that input. Once the character is created, the role of AI largely ends. During gameplay, the AI does not attempt to replace the player.

After the demo, I had a brief conversation with a developer from 10six Games. He explained that aside from using generative AI for visual assets, most of the game was built by human developers. However, I raised a concern. Since players can modify their characters through prompts, is it possible for a “solution prompt” to emerge? In other words, could players discover an optimal prompt that guarantees the strongest possible build, thereby breaking the game’s balance? Before the rise of AI, game designers controlled balance directly. If that control is partially handed over to players, how can we ensure that the overall player experience remains intact? He acknowledged that this is indeed a potential issue and admitted that there is currently no perfect solution. Their approach, for now, is to encourage players to experiment with different prompts and find enjoyment in the process.

AI can generate content, assist production, and even automate certain systems, but there are still many aspects of game design that it cannot replace. Game design is about creating meaningful interactions, shaping player decisions, and building systems that respond to those decisions in intentional ways. These are not simply problems of generation, but of judgment.

The experiences I had at GDC made it clear that the problem is not whether AI can be used in games, but how it is used. When AI begins to replace player agency or design intent, it risks weakening the very foundation of what makes games engaging. On the other hand, when it is used as a tool to support human creativity, it can expand what designers are able to create without taking control away from the player. At least for now, AI remains a powerful tool rather than a designer. Making a game faster or more complex does not necessarily make it more meaningful. Ultimately, what makes a game engaging still depends on human decisions, not machine generation.