The conversation about AI in game development usually centres on art and writing — the visible, emotional outputs that generate the most debate. But a quieter adoption is happening in game design itself, where AI tools are being used to assist with systems design, level layout, and gameplay balancing.

I’ve spent the last few months talking to designers at Australian studios about how they’re using these tools. The picture is more nuanced than either the hype or the backlash suggests.

What designers are actually using

The tools fall into three broad categories:

Brainstorming and ideation assistants. LLM-based tools (mostly custom GPT configurations) that designers use to generate ideas, explore design space, and work through systems problems. Think of it as a conversation partner that never gets tired of discussing combat balance at 11pm.

Level design assistance. Tools that can generate layout variations based on parameters — room size, enemy placement density, objective locations. The designer defines the constraints, the AI generates options, the designer picks and refines.

Balance and economy modelling. AI systems that simulate thousands of playthroughs to test game economy systems, difficulty curves, and progression pacing. These overlap with the testing tools I’ve written about before, but they’re being used earlier in the design process rather than as a QA step.

What designers told me

A lead designer at a Melbourne studio has been using LLM assistants for brainstorming for about six months. “It’s like having a junior designer who’s read every game design book ever written but has never actually played a game. The suggestions are technically informed but lack practical intuition. About one in ten ideas is genuinely useful. The rest are variations of things that already exist.”

A level designer at a Sydney studio tested AI layout generation and found it useful for creating initial blockouts — rough layouts that serve as starting points for detailed design. “It saves me maybe two hours per level on the blockout phase. I still have to redesign 60 to 70 percent of what it produces, but having a starting point that’s roughly in the right ballpark is better than starting from a blank canvas.”

A systems designer at a Brisbane studio uses AI economy modelling extensively. “This is where I’ve seen the clearest value. We can test 10,000 variations of a resource economy in an afternoon. Without AI, that testing would take weeks of manual simulation. The tool doesn’t tell us what the economy should be — it tells us which of our ideas break and which ones hold up.”

Teams that have worked with AI agent builders Sydney to build custom design tools have found that purpose-built AI assistants perform significantly better than general-purpose LLMs for specific design tasks.

The creative homogeneity concern

The most thoughtful criticism I heard came from an independent designer in Melbourne. “When everyone’s using the same AI tools with the same training data, there’s a risk that the design space converges. The AI is drawing from the same pool of existing games and design patterns. If everyone’s brainstorming with the same tool, everyone’s going to arrive at similar ideas.”

This is a legitimate concern. The strength of game design as a creative discipline is its diversity — different designers bring different influences, experiences, and perspectives. If AI tools push everyone toward the median of existing design, the industry loses the outlier ideas that produce the most interesting games.

The counterargument is that designers are using AI as a starting point, not an endpoint. A good designer takes an AI suggestion and transforms it through their own creative lens. But the concern about convergence is worth monitoring as adoption increases.

Where it’s most useful

Based on my conversations, AI design assistants provide the most value in:

Early-stage iteration. When a designer needs to explore a large design space quickly — testing many variations of a system before committing to one — AI dramatically accelerates the process.

Documentation and communication. AI can help translate design intuition into clear documentation. A designer who knows what they want but struggles to articulate it in a design document can use AI to structure their thinking.

Player behaviour prediction. AI models trained on player data can predict how players will interact with new systems. This doesn’t replace playtesting, but it identifies likely problems earlier in the process.

Where it’s least useful

Novel mechanics. AI tools are trained on existing games. Genuinely new gameplay mechanics — the kind that define generations of games — don’t come from pattern matching on existing data. They come from creative leaps that AI currently can’t make.

Emotional design. The moments in games that make players feel something — joy, fear, surprise, sadness — are the product of intentional creative decisions. AI can generate technically competent design. It can’t generate design that moves people.

Cultural specificity. An AI trained primarily on Western and Japanese game design will produce suggestions rooted in those traditions. Australian-specific design sensibilities, references, and cultural context need human input.

My take

AI game design assistants are useful tools. They’re not creative partners, they’re not going to replace designers, and they’re not going to ruin game design. They’re productivity aids that handle some of the tedious parts of the design process faster than humans can.

The studios that use them well will ship better games faster. The studios that over-rely on them will ship games that feel generic and derivative. The tool itself is neutral. The decisions about how to use it are what matter.