When AI can make a game, what is left for the designer?

SKIP TO CONTENT

Tiếng Việt

When AI can make a game, what is left for the designer?

The more generative AI can automate game development, the more the industry's success will depend on distinctly human qualities.

Concern about generative AI is rising even as its use becomes more widespread. The Game Developers Conference’s 2026 State of the Game Industry report found that 52 per cent of developers believe the technology is negatively affecting the games industry, up from 30 per cent in 2025 and 18 per cent in 2024. Meanwhile, a separate survey found that more than a third of game workers (36 per cent) already use AI in their work.

The real-world stakes of this shift became clear when the highly anticipated indie game Clair Obscur: Expedition 33 was stripped of its Game of the Year and Best Debut Game titles at The Indie Game Awards after the organisers discovered the developers had used unauthorised generative AI assets during production.

These developments capture a growing tension at the heart of game development. Generative AI is increasingly embedded in everyday workflows, from research and coding to prototyping, yet many developers remain concerned about its impact on the industry, particularly when it touches the expressive heart of games: art, music, narrative and character.

It is tempting for universities to adopt a simple position: either embrace AI as an inevitable professional tool or prohibit it to protect learning and authorship. Neither response prepares students for the industry they are entering.

The more useful educational question is not simply whether a student used AI. It is what they delegated, why they delegated it, how they evaluated the result and whether they remained responsible for the work.

The real question is not whether students use AI, but how they use it. (Image: RMIT)

Mr Pham Hong Nhat, Associate Lecturer in Game Design program at RMIT Vietnam, said: "In our game development teaching at RMIT Vietnam, we have explored this question at two different stages of creative practice.

"During pre-production, we have used AI as a Socratic interrogator rather than an idea generator. Students bring their own concepts, references and intentions. The tool is asked to probe assumptions: Why should this environment exist? What does this visual choice communicate? What cultural material is being used, and has it been researched carefully? What consequence follows from this mechanic?"

Rather than allowing AI to become a shortcut around thinking, this approach uses the technology to challenge students beyond shallow thinking. When a student’s proposal is vague, the system can ask “why?” on almost every line. The student must still decide which questions matter and what the answers mean.

During production, lecturers in RMIT Vietnam’s Game Design program have adopted a different approach. Mr Dodick Zulaimi Sudirman explained that in Game Design Studio 5 course, students were introduced to a documented workflow for code assistance: Specify, Plan, Execute and Verify.

"The specification and plan are not disposable prompts; they are evidence of the student’s intention. Execution is bounded, and verification is a required act rather than an optimistic glance at whether the game launches" he added.

The workflow places four responsibilities back on the human creator: delegation, deciding what can appropriately be handed over; description, expressing the desired behaviour precisely; discernment, judging whether the output is correct, safe and appropriate; and diligence, testing, disclosing and reflecting on the process.

These responsibilities matter because AI does not remove creative difficulty, instead relocating it.

When code, images and concepts can be produced quickly, making more is no longer the same as making well. The scarce capabilities become judgment, taste, context and the willingness to stay with a difficult idea long enough to understand it. AI can accelerate production, but it cannot supply a compelling reason for a game to exist.

Students may draw different boundaries around AI. (Image: RMIT)

Experience in RMIT Vietnam's Game Design classrooms suggests that critical AI education does not require every student to reach the same conclusion. In one small class, most students accepted AI assistance for code while drawing a boundary around art, music and story. One was comfortable going further; another declined to use it on principle. The process did not convert them to a single institutional position. It required them to articulate and test their own positions.

Dr Renusha Athugala, Program Manager for Game Design at RMIT Vietnam, said: "That distinction is important. Teaching students to use AI critically should not become a polite way of compelling adoption. Refusal can be a legitimate design and ethical decision. So can carefully bounded use. What matters is that the decision is informed, visible and answerable."

As generative systems grow more capable, educators should demand more from students, not less. The focus should be on clearer intent, stronger research, more ambitious experimentation and rigorous verification. Learning is better measured by the quality of human judgement surrounding AI use than by the presence or absence of the technology itself. For future game developers, technical fluency will matter, but the defining skill may be knowing what not to automate and why.

Story: June Pham

Communication and design

Media Release

15 October 2026

Share

Engage

Contact RMIT News

Find an expert

RMIT Vietnam facts & figures

Related news