Courtesy of Smilegate

If a newly released game shows that it used artificial intelligence (AI), does that make it a well-made game? Debate is intensifying over how much AI use is appropriate in the game development and production process. The industry says AI adoption is inevitable given expense savings and shorter production schedules, but players say noticeable AI use breaks immersion and are pushing back.

Clair Obscur: Expedition 33 by French startup Sandfall Interactive, cited as last year's most talked-about title, was stripped last month of its Indie Game Awards (IGA) game of the year prize for using AI in production. IGA requires at the nomination stage that generative AI not be used in the production process, and determined the game violated that rule. The group said AI use runs counter to the indie game scene's values of pursuing pure creation.

After the prize was rescinded, online communities were split. Some fans argued that because Expedition 33 won praise for its distinctive worldbuilding, stylish direction, high aesthetic completion, and gameplay, limited AI use in production should not be an issue. The game in fact swept nine awards, including the top prize, at The Game Awards (TGA) 2025, the industry's most prestigious awards. Others said that given a game is a single work that mobilizes a studio's full creative capabilities, they support IGA's policy of evaluating and awarding only games that exclude AI use. As the controversy grew, Steam, the world's largest game platform, tightened its rules to require strict disclosure of whether AI assets were used.

Separate from the authorship controversy, a dominant view is that AI adoption in gaming is an irreversible trend. Studios at home and abroad are rushing to introduce AI to boost development efficiency and cut expenses. AI automates repetitive tasks to ease developers' burdens and is also used to make gameplay more varied.

ReLU Games, a development studio under Krafton, is actively adopting AI in production to build polished games with small teams. Magical Girl Luluping, which ReLU Games unveiled in 2024, drew attention after a three-person team used Generative AI to build an internal demo in a month. Aside from some special effects and sounds, the studio said 90% of the game's assets (components) were made with AI. Considering that even the simplest casual mobile game takes at least two to four months to build by hand, AI sharply cut the production time.

Mimesis, ReLU Games' new title that recently topped 1 million in cumulative sales, is a representative case applying reinforcement learning and small language model (SLM) technology. In the game, AI mimics a user's behavior and voice and appears as an enemy disguised as a teammate, which reviewers said boosts fun and tension.

Courtesy of Krafton

ARC Raiders, a hit developed by Nexon subsidiary Embark Studios, also used AI for in-game NPCs (non-player characters) and in-game voice production. NCSOFT is applying Barco, a large language model (LLM) developed by subsidiary NC AI, across game development. NC AI's Barco 3D cuts 3D game asset production, which used to take more than three weeks, to around 10 minutes.

Some analysts say that as the global game industry grapples with intensifying competition, rising development costs, and the rise of shorts and other substitute content, AI adoption has already become a necessity, not a choice. The argument is gaining traction that smaller indie studios with limited budgets, in particular, should use AI to raise quality and create more experimental, differentiated games. According to a survey Google Cloud released last year, 87% of 615 developers in five countries — Korea, the United States, Norway, Finland, and Sweden — said they already use AI in game development.

The trend shows up in game statistics as well. A report last year by Totally Human Media found that of roughly 114,000 games on Steam, 7% (7,818 titles) used Generative AI, and among new releases the share was higher at 20%. That was an increase of about 700% from the previous year.

In an interview with a Japanese outlet in Nov. last year, Lee Jeong-heon, head of Nexon's Japan unit, said, "All game companies are already using similar AI technologies," and added, "The question is how to use AI to build a competitive edge." Tim Sweeney, chief executive officer (CEO) of Epic Games, Inc., also said, "It no longer makes sense to categorize games based on whether they use AI."

But players still resist having AI replace core creative areas like character design, illustration, and story. For example, the latest installment of the popular shooter Call of Duty was engulfed in controversy for using AI images. If the AI images had blended naturally into the game, there would have been no issue, but the title included a number of low-quality "AI slop," including the "Ghibli-style images" that trended on ChatGPT, drawing criticism that it "deceives players."

The voice AI used in ARC Raiders also drew reviews that "its robotic voices break immersion." The developer explained, "We used AI to cut the expense and time spent on repetitive line production, and we did not apply AI to visual elements."

The industry expects that as AI technology advances and the quality of AI-generated content improves, the uncanny valley effect typical of AI-made works will gradually be resolved. According to market research firm Grand View Research, the game AI market, which was $3.3 billion (about 4.8 trillion won) last year, is projected to grow to $51.3 billion (about 74 trillion won) by 2033.

※ This article has been translated by AI. Share your feedback here.