Nintendo Said No to Generative AI. 85.8% of Japan’s Game Developers Said Yes

Two years ago, Nintendo told investors it would not use generative AI to make its games, citing worries about intellectual property. It was the most famous studio in the world drawing a line.

This week, the rest of Japan’s game industry showed where it actually stands. According to a preview of the CESA Video Game Industry Report 2026, published on September 17 ahead of Tokyo Game Show, 85.8 percent of Japanese game developers now use generative AI in their work. Last year the figure was 51 percent.

A 34 point jump in twelve months is not a trend. It is a switch being flipped. Below is what the number means, what developers are doing with the tools, why studios say they need them, what is worrying the people who run the industry, and what happens to the artists whose jobs sit closest to the machine.

One number, four layers

The 85.8 percent is a headline, and headlines hide detail. The survey behind it, titled Game Developers’ Employment and Career Formation 2026, collected 1,349 valid responses between May and August from directors, sound staff, technical artists, executives, rank and file employees and students. Broken down:

Two of those rows matter more than the headline. Nearly two thirds of Japanese game developers now touch generative AI every working day, which makes it closer to a spreadsheet than a novelty. And fewer than one in twenty have never tried it. CESA’s own reading is that the tools have moved past early adopters and are now embedded across development sites.

It is also worth noting how the baseline shifted. The 2025 figure of 51 percent came from a survey of 54 companies. The 2026 figure comes from individual developers. The two are not perfectly comparable, but every independent data point over the past year has pointed the same direction, so the jump is real even if the exact size is soft.

What studios say they get out of it

CESA ran a second, company level survey alongside the developer one. When firms were asked why they had brought generative AI in, the answers were about business pressure, not creative curiosity:

• Shorter development cycles (30 companies)
• Lower development and operating costs (29)
• Multilingual support and faster global launches (24)
• New forms of expression and ideas (22)
• Better content and service quality (21)
• Dealing with labour shortages (20)

Read that list against Japan’s situation. The country has an ageing workforce, a chronic shortage of engineers, a weak yen that makes overseas revenue vital, and a games sector whose biggest hits depend on selling in dozens of languages. Generative AI is being adopted as an answer to those problems first and as an art tool second.

Which tools, and for what

A separate survey of Japanese online game companies, reported by Automaton in July, found adoption in that segment had reached 100 percent. It also asked which tools developers reached for. Google’s Gemini was used by 94 percent of respondents, Anthropic’s Claude by 84 percent and GitHub Copilot by 76 percent. The most common use case was not art or voice but analysing what players want, followed by code, translation and text generation.

The 2025 CESA survey had ranked image and visual asset generation as the top use, with story and text second and programming support third. A year later, programming and analysis have overtaken art in most reporting. That shift is partly about capability, since coding models improved sharply in 2026, and partly about risk, since generated code raises fewer public relations problems than generated character art.

The guardrails studios are building

CESA’s company survey also asked how firms manage the tools, and the answers show an industry that has learned from the backlash of the past two years.

The most common practice was that humans handle verification, correction and supervision of anything AI produces. A majority of companies also restrict or specify which tools staff may use, and a majority avoid putting generated output into products as is. CESA’s interpretation is that studios are chasing productivity while building a human centred operating framework around it.

This is the difference between Japan’s 2026 numbers and the free for all of 2023. Adoption went up because the rules went in.

Why the man announcing the number sounded cautious

CESA Executive Director Tsutomu Masuda did not celebrate the 85.8 percent. His statement said generative AI is being adopted widely and steadily, then immediately noted that concerns about rights, including copyright infringement, have been raised and require care. He framed the industry’s task as balancing the possibilities of the technology with rights protection and the quality of human centred development.

He has reasons to be careful. In September a Shanghai court ordered an AI voice service to pay Genshin Impact’s developer $112,000 for cloning 63 character voices, a reminder that AI misuse cuts both ways for game companies. Evangelion voice actress Megumi Ogata said this month she was heartbroken by unauthorised AI versions of her voice and asked why Japan has no content guidelines. And in August, a Japanese government panel approved a principle code that would push generative AI companies to disclose their training data, the first sign that Tokyo’s famously permissive AI copyright regime may be tightening.

For an industry whose most valuable assets are characters, Masuda’s caution is not hedging. It is risk management.

Level-5, Nintendo and the two roads

The week’s smaller stories show the split inside Japan.

Level-5, maker of Professor Layton and Yo-kai Watch, had its CEO acknowledge in a presentation this week that the company uses generative AI. That matters because Level-5 spent part of 2026 dealing with fan accusations that AI generated art had slipped into promotional material. The company has gone from denying suspected use to openly confirming it, which is roughly the journey the whole industry has taken.

Nintendo, by contrast, has held its position that it will not use generative AI in game development because of IP concerns, and former Nintendo art director Takaya Imamura has publicly experimented with the tools only to voice doubts about them. Nintendo is a CESA member. Its stance is now the exception in its own trade body.

Neither road is obviously wrong. Nintendo’s brand is built on a house style so distinctive that any hint of machine generation would be a headline. A mid sized studio shipping a live service game in nine languages has a different calculation.

The people closest to the machine

Adoption numbers are abstract until you ask who does the work the AI now does. Automaton’s reporting on the Japanese industry names the roles most affected: background and prop artists, and assistant programmers.

Those are entry level positions. They are how artists and coders traditionally learned the craft, by drawing the hundredth rock or fixing the hundredth bug before being trusted with anything important. If generative AI absorbs that work, the industry saves money now and faces a question later: where does the next generation of lead artists and senior engineers come from?

This is the same tension universities are confronting, as we covered in our piece on AI disruption in academia. The tool does the beginner’s job, and the beginner never becomes an expert. Japan’s studios, with their labour shortage, may feel that trade off later than most.

How Japan compares to everyone else

Japan is not following the global curve. It is ahead of it.

The global GDC survey found only 36 percent of game professionals personally use generative AI, and about half of companies use it anywhere in the organisation. Unity’s 2025 report found 79 percent of developers feel positive about the technology, which measures attitude rather than use, but suggests the appetite exists everywhere. Japan simply converted the appetite into practice faster, and the CESA numbers imply South Korea and Southeast Asia will report similar jumps within a year.

What this means if you play games, make them or invest in them

Players: the next wave of Japanese games, including titles announced at Tokyo Game Show this week, was largely built with AI in the pipeline. Most of it will be invisible: faster localisation, more polished code, analytics that shape balance patches. The visible risk is art or voice that feels off, and the CESA data suggests studios know that and are keeping humans in the loop precisely to avoid it.

Developers: the 2026 survey is a job market signal. Daily fluency with generative tools is now assumed at most Japanese studios. The roles under most pressure are the ones the tools do best, and the roles gaining value are verification, direction and the judgement to know when a generated asset is wrong.

The industry: Japan has run the experiment the rest of the world is still debating. Its answer so far is that adoption and guardrails can rise together. Whether that holds when the first major copyright case lands against a Japanese studio, rather than in its favour, is the question Masuda was really raising.

Frequently Asked Questions