Survey Finds 85.8% of Japanese Game Developers Now Use Generative AI

Generative AI use appears to be accelerating rapidly across Japan's game-development sector. A preliminary summary from the Computer Entertainment Supplier's Association, or CESA, says that 85.8% of Japanese game developers who responded to its recent survey are actively using generative AI as part of their work.

The finding represents a major increase from the previous year's result, when 51% of surveyed developers reported using the technology. The new survey was conducted among developers attending CEDEC, Japan's long-running game developer conference, and received responses from 1,349 people. CESA plans to publish its complete report in December.

That headline figure does not mean every responding developer is using AI to create art, characters, or other content that appears in finished games. The survey covers a broad range of uses, from generative tools supporting day-to-day office tasks to assistance with development workflows. Still, the results show that AI has moved far beyond a niche experiment for many professionals in the Japanese games business.

AI adoption is broad, but its uses differ

The respondents described using generative AI in both production and administrative contexts. Tools such as ChatGPT and Copilot can help with tasks including writing, research, documentation, code assistance, troubleshooting, and debugging. The technology is also being used in some cases to generate game-development assets.

For useful background on this topic, read GTA 6 May Skip Florida's Biggest Theme Parks as Rockstar Focuses on Leonida's Identity.

For studios and individual developers, those categories carry very different implications. A tool used to summarize internal notes or help identify a programming issue may never affect a game's visible creative output. By contrast, AI-generated visual material, writing, localization, and player-facing content can become a far more contentious subject, particularly when audiences cannot tell how or where a tool was used.

The CESA results suggest frequent use is already normal for a large segment of respondents. Among developers using the technology, 63% said they use it every day. Another 22.8% said they use it occasionally. Only 8.6% said their companies had introduced guidance covering AI use, though a separate company-focused portion of the research paints a fuller picture of policy-making at the studio level.

Daily usage does not necessarily indicate a single, standardized workflow. One developer may rely on an AI assistant while writing code, while another may use it for draft documentation or idea organization. Some teams may employ it during pre-production, while others may limit it to technical support. That distinction matters when interpreting the survey: high adoption indicates that the tools are present in developers' routines, but it does not establish that they are being used in the same way across projects or companies.

Developers see gains in speed, cost, and accessibility

Most respondents who use generative AI said they had seen benefits. Many pointed to reduced production time and lower costs, two incentives that are particularly powerful in an industry where projects can require huge teams, lengthy schedules, and repeated rounds of iteration.

CESA's preliminary findings also indicate that 837 respondents believed the technology expanded creativity. A further 738 said it lowered technical barriers. Those answers reflect one of the most common arguments for AI-assisted tools: they can potentially help a developer explore options more quickly, handle routine work, or attempt tasks that might otherwise demand specialized knowledge.

In game development, that could mean helping programmers investigate an error, allowing a designer to rapidly organize prototype ideas, or speeding up the creation of internal reference material. The survey does not provide a complete task-by-task accounting of how each respondent used AI, so it cannot determine which specific applications delivered the strongest results. It does, however, show that many developers believe the tools have practical value within their own work.

There is an important contrast in the data. Only 147 respondents said generative AI benefited the end-user experience. That gap suggests developers may currently see AI as more useful behind the scenes than as a direct way to make games better for players.

That result should not be surprising. Saving time on a repetitive task can help a development team, but it does not automatically produce better art direction, better level design, stronger writing, smoother performance, or a more engaging game. The value of a game remains tied to execution, creative judgment, and careful testing. AI may assist with pieces of that process, but players ultimately judge the finished product rather than the efficiency of the production pipeline.

Company policies reveal a more cautious picture

The developer survey has limitations that are important to keep in mind. Because it collects answers from individual professionals, multiple respondents may work for the same company and follow identical policies. A studio with a large number of respondents could therefore have a greater effect on the total than a smaller studio, even if both businesses take a similar approach to AI.

CESA also surveyed member companies separately. In that portion of the research, 74% of companies said they had policies governing generative AI use. That finding suggests studios are not simply allowing unrestricted adoption. Instead, many are establishing rules about where the technology can be used, what material can be entered into third-party systems, and which outputs require review.

The company responses were also much more restrained than the individual developer figure might initially imply. There was a nearly even divide between companies that were not using generative AI and those using it in some fashion: 27 companies said they were not using it, while 26 said they were.

Only 12 companies reported using generative AI for visual assets. Fewer still said they used it for areas such as story work, localization, quality assurance, or programming. Those results reinforce the idea that widespread familiarity with AI tools does not automatically translate into widespread use in every stage of commercial game production.

A company may permit developers to use AI for limited internal productivity work while prohibiting it from being used to create final assets. Another may allow experimentation but require legal, technical, or creative approval before any output is included in a project. Such distinctions are increasingly relevant as studios attempt to balance perceived productivity benefits with questions around quality control, intellectual property, data security, labor, and public perception.

Player reaction remains a major factor

Generative AI has become one of the most divisive subjects in games. Players have raised concerns about whether visual assets are authentic, whether creative work has been replaced rather than supported, and whether companies are using automation primarily to cut costs or reduce jobs. Those concerns are especially intense when AI-generated material is suspected in prominent art, promotional material, or finished game content.

Recent attention around Level-5, the developer behind Dark Cloud and Professor Layton, illustrates the scrutiny that can follow. Fans suspected that a presentation from the company included a substantial amount of AI-generated imagery. Regardless of the broader uses studios may find for AI, cases like that demonstrate why player trust can be affected by the way tools are implemented and communicated.

Japan and South Korea have generally shown stronger support for these technologies than some other regions, but that does not remove the risk of backlash. Global reactions to AI-related decisions in games, including criticism surrounding Crazy Taxi: World Tour, show that audiences remain ready to challenge companies they believe are using the technology carelessly.

For publishers and developers, the practical lesson may be that AI policy cannot be treated solely as an internal efficiency issue. A studio could see meaningful benefits from code support, documentation help, or debugging assistance while still facing a negative response if its player-facing use of AI appears unconsidered or obscured. Clear boundaries and thoughtful communication may matter as much as the technology itself.

What the survey says about the road ahead

CESA's preliminary data points to a Japanese development community that is rapidly incorporating generative AI into its working habits. The jump from 51% to 85.8% in a year is substantial, even with the survey's respondent-based limitations. It indicates that developers are increasingly encountering these tools, testing them, and in many cases incorporating them into regular routines.

At the same time, the company data shows a sector still setting limits. Visual assets are not the dominant reported use among surveyed companies, and use in story, localization, QA, and programming remains less common. The difference between personal usage and official studio policy may become one of the most important aspects of AI adoption going forward.

For now, the strongest reported benefits appear to concern developers' own workflows: faster production, reduced cost, broader experimentation, and lower barriers to certain tasks. The much smaller number of respondents who saw direct benefits for players is a reminder that adoption alone is not a measure of creative success.

As CESA prepares its full December report, the most significant question may not be whether generative AI will remain part of game development. The preliminary findings strongly suggest that it will. The more consequential issue is how studios define acceptable use, protect the work and data involved in development, and ensure that productivity tools do not come at the expense of the confidence players place in the games they buy.