Japan’s game industry just put a hard number on something developers have been saying anecdotally for two years: generative AI has stopped being a side experiment and become part of the daily job. The Computer Entertainment Supplier’s Association, known as CESA, released a preview of its 2026 Video Game Industry Report at Tokyo Game Show on September 17, 2026, and the headline figure is stark. 85.8% of Japanese game developers now use generative AI in their work, with 63.0% doing so daily and another 22.8% using it occasionally, according to AUTOMATON, which cited the Japanese outlet Denfaminicogamer.
That figure lands at a sensitive moment for the industry. Studios are under pressure to ship faster with smaller teams, publishers are watching AI-generated art controversies play out in public, and CESA’s own member roster includes some of the biggest names in gaming: Capcom, Sega, Level-5, and Konami. What the survey shows is not a hypothetical future. It shows a workforce that has already crossed the adoption threshold, and it raises a harder question about what comes after adoption, namely quality control, copyright exposure, and whether Japan’s shift outpaces the rest of the world.
What CESA’s 2026 Survey Actually Found
CESA’s preview report is drawn from a survey of game developers and CESA member companies, collected between May and August 2026 and compiled into 1,349 valid responses. The full 2026 CESA Video Game Industry Report is scheduled for release in early December 2026, but the organization pulled the generative AI figures forward for its Tokyo Game Show preview because the numbers were considered significant enough to headline on their own, according to reporting from AUTOMATON and the outlet Yomimono.
The topline number, 85.8%, combines two groups: developers who use generative AI every day and developers who reach for it only sometimes. CESA did not, in this preview release, break out which specific tools or models developers are using, and it did not disclose a studio-by-study breakdown. What it did share is a set of benefits developers say they get from the technology, and a set of controls studios say they’ve put in place to manage the risk. Both of those lists matter more than the topline percentage, because they tell you what the adoption is actually for.
Daily Habit, Not Just a Tool: Breaking Down the 85.8%
Splitting the 85.8% into its two components changes the story. A tool that 22.8% of workers touch occasionally is still a notable adoption rate, but it reads as experimental. A tool that 63.0% of workers touch every single day reads as infrastructure. CESA’s numbers put Japanese game development firmly in the second category. Generative AI, per this survey, is not a novelty plug-in that a handful of enthusiasts use to speed up concept art. It is a default part of the workflow for roughly two out of every three developers surveyed.
That distinction matters for how studios plan training, tooling budgets, and legal review processes going forward. A once-in-a-while habit can be managed with loose guidelines. A daily habit needs a formal policy, version control for AI-touched assets, and a clear chain of accountability when a generated asset turns out to resemble copyrighted material too closely. CESA’s survey suggests most studios have already had to build some version of that policy, whether or not it is written down.
The CESA 2026 Survey at a Glance
| Metric | Figure | Detail |
|---|---|---|
| Total generative AI usage | 85.8% | Combines daily and occasional use |
| Daily use | 63.0% | Use generative AI as part of routine work |
| Occasional use | 22.8% | Use generative AI intermittently |
| Survey sample | 1,349 responses | Valid responses from developers and CESA member companies |
| Collection window | May–August 2026 | Fieldwork period ahead of the preview release |
| Preview release date | September 17, 2026 | Announced at Tokyo Game Show 2026 |
| 2025 member-company usage | ~51% | Prior-year figure, company-level not developer-level |
| 2025 in-house engine AI use | 32% | Share of companies using AI to help build proprietary engines |
| Full report release | Early December 2026 | Complete 2026 CESA Video Game Industry Report |
Who CESA Surveyed: Capcom, Sega, Level-5, and Konami
CESA’s member roster is not a random sample of small studios. It includes a large number of prominent Japanese developers and publishers, among them Capcom, Sega, Level-5, and Konami, according to AUTOMATON’s reporting. That matters for how much weight the 85.8% figure should carry. This is not a survey of hobbyist developers or indie teams experimenting with AI on the side. It’s a signal from companies that ship AAA and mid-size titles at scale, the kind of studios whose production pipelines set norms that smaller Japanese studios often follow.
When a company the size of Capcom or Sega adopts a workflow change, it tends to ripple outward through contractors, outsourcing partners, and the broader supplier ecosystem that CESA itself represents. CESA, after all, is the Computer Entertainment Supplier’s Association, the organization that runs Tokyo Game Show and represents the commercial interests of Japan’s game hardware and software suppliers. A survey run by that body, targeting its own membership, is effectively a temperature check on the companies that decide how the rest of the industry works.
The Case for Adoption: Efficiency, Speed, and Lower Costs
CESA asked member companies what they expect to gain from generative AI, and the answers followed a predictable but telling order. The most commonly cited benefit was improved operational efficiency and productivity, followed by shorter development cycles, and then reduced development and operational costs, according to AUTOMATON and Yomimono’s reporting on the same preview data. None of those three answers are about creative capability. They’re about throughput.
That framing lines up with how Japanese publishers have talked about production costs over the past few years. Development budgets for console and PC titles have climbed steadily as visual fidelity and scope have grown, while headcount growth hasn’t kept pace at most studios. Generative AI, in this context, functions less like a creative collaborator and more like a cost-control lever, something that lets a fixed-size team produce more assets, iterate on more concepts, and cut turnaround time on repetitive tasks such as background art, placeholder dialogue, and early-stage prototyping.
From 51% to 85.8%: How Fast Did This Move?
AUTOMATON’s headline framed the new figure as an increase from last year’s 51%, and that comparison is worth examining carefully rather than repeating at face value. CESA’s 2025 Video Game Industry Report measured generative AI use at roughly 51% among CESA member companies, a company-level statistic drawn from 54 Japanese game companies. This year’s 85.8% figure measures something narrower and more granular: individual developer usage, drawn from 1,349 valid survey responses. A company reporting that it uses generative AI somewhere in its pipeline is a different measurement than 85.8% of individual developers reporting personal use of the technology.
That methodological gap doesn’t make the 2026 number less significant. If anything, a developer-level survey showing 85.8% personal usage is arguably a stronger signal of on-the-ground adoption than a company-level yes/no answer. But treating the two figures as a clean, apples-to-apples “34-point jump” oversells the precision of the comparison. The honest read is that generative AI use in Japanese game development went from a majority practice among companies in 2025 to a near-universal daily or occasional habit among individual developers in 2026, measured two different ways.
The 2025 report also found that 32% of Japanese game companies used AI to help develop their own in-house game engines, and that the most common uses of AI at the time were generating visual assets and images, followed by story and text generation, and then programming support. CESA’s 2026 preview didn’t restate a fresh version of that specific breakdown, but nothing in the new data suggests those use cases have shrunk. If anything, the jump in daily usage implies they’ve deepened.
The Guardrails: Human Review Over Outright Bans
CESA also asked developers what measures they have in place when using generative AI, and the most common answer was conducting human verification, correction, and supervision of AI output. A large share of respondents also said they specify or restrict which tools they’re allowed to use, and many said they avoid using generated output directly, treating it instead as a draft or starting point rather than a finished asset, per AUTOMATON’s account of the preview data.
That’s a notably different posture from the blanket bans some Western studios and publishers floated in 2023 and 2024 amid union pushback and public backlash over AI-generated art. Japan’s largest studios appear to be converging on a middle path: keep the tools, but insist a human touches every output before it ships. Whether that policy holds up under production pressure, especially near a launch deadline when review time gets squeezed first, is an open question CESA’s survey doesn’t answer.
Where Generative AI Fits in the Development Pipeline
Even without a fresh 2026 breakdown by task, the pattern from CESA’s 2025 data plus this year’s expected-benefits ranking gives a reasonable picture of where the technology sits inside a Japanese studio’s pipeline. Visual asset generation, including background art, texture variations, and early concept passes, remains the most cited use. Text generation, covering dialogue drafts, item descriptions, and localization support, comes next. Programming support, from boilerplate code to debugging assistance, rounds out the top three, a pattern that echoes what Western engine vendors and coding-assistant makers have reported among their own developer bases.
Not every corner of the industry has embraced this quietly. Earlier in 2026, Level-5 had to publicly address online scrutiny after fans flagged what looked like AI-generated artifacts in one of its promotional trailers, a controversy the studio said stemmed from human production errors rather than careless AI use, according to AUTOMATON’s coverage. Separately, Capcom engineers have said publicly that even sophisticated AI “can’t match our creators when it comes to sensibility,” a reminder that adoption numbers and creative confidence in the tools are two different things.
Japan vs the World: How This Compares Globally
Set against the rest of the games industry, Japan’s 85.8% figure looks unusually high, though the comparison again requires some care because the surveys don’t ask identical questions. The Game Developers Conference’s State of the Game Industry survey, one of the most cited industry-wide polls, found that 36% of game industry professionals personally use generative AI tools as part of their job, while 52% said their company uses generative AI tools somewhere in the organization. South Korea’s game industry, tracked separately by the Korea Creative Content Agency, reported a generative AI utilization rate of 41.7% in the first half of 2025.
Sentiment data adds another layer. Unity’s 2025 Gaming Report found that 79% of developers feel positive about generative AI, a figure that measures attitude rather than actual usage, but it suggests the appetite for adoption exists well beyond Japan even where usage numbers lag. Read together, these numbers suggest Japan’s game development sector, at least among CESA’s member companies, has moved further and faster on daily generative AI use than the global average reported by GDC or the regional figure reported for South Korea.
Global Generative AI Adoption in Game Development
| Region / Survey | Metric | Figure |
|---|---|---|
| Japan, CESA 2026 (developers) | Any generative AI use | 85.8% |
| Japan, CESA 2026 (developers) | Daily use | 63.0% |
| Japan, CESA 2025 (member companies) | Any generative AI use | ~51% |
| Global, GDC State of the Game Industry | Personal use by professionals | 36% |
| Global, GDC State of the Game Industry | Company-level use | 52% |
| South Korea, KOCCA (H1 2025) | Industry utilization rate | 41.7% |
| Global, Unity Gaming Report 2025 | Positive sentiment toward generative AI (not usage) | 79% |
Historical Context: Japan’s Cautious Turn Toward Generative AI
Japan’s relationship with generative AI in creative industries has not been a straight line toward enthusiastic adoption. The country’s game and animation sectors spent much of 2023 and 2024 debating AI-generated art in public, with fan communities frequently calling out suspected AI use in trailers and marketing material, sometimes correctly and sometimes not, as the Level-5 incident referenced above illustrates. That scrutiny pushed some studios toward caution and disclosure policies rather than a rush to automate visible creative work.
What CESA’s numbers suggest is that the caution around visible, customer-facing AI use didn’t translate into caution about internal, production-side AI use. Developers kept adopting the tools for the unglamorous parts of the job, prototyping, drafting, debugging, and asset iteration, even as public debate raged over whether AI belonged anywhere near a finished game. The 2025-to-2026 shift CESA is reporting looks less like a sudden change of heart and more like an internal habit that had already been building for a couple of years, only now getting measured and published at scale.
Market Impact: What This Means for Studios and Investors
For publicly traded Japanese publishers such as Capcom, Sega Sammy, and Konami, a survey confirming near-universal internal AI adoption is a data point investors will likely fold into cost-structure expectations. If generative AI genuinely shortens development cycles and reduces operational costs, as CESA’s respondents expect, the effect should eventually show up in either faster release cadences, higher margins on existing headcount, or both. None of that has been demonstrated in a single quarterly filing tied specifically to this survey, and CESA’s report doesn’t attempt to quantify a dollar figure or percentage cost reduction, so any margin impact remains an expectation rather than a booked result.
There’s also a competitive dimension. Outsourcing and contract studios that support CESA’s larger members, the background art houses, localization vendors, and QA contractors that make up much of Japan’s game production supply chain, will likely face pressure to match the AI-assisted throughput their clients now expect. A vendor still working at pre-AI production speed risks looking slow and expensive next to one that has already integrated generative tools into its own pipeline, regardless of whether the client studio explicitly requires it.
The Copyright Question CESA Hasn’t Answered
CESA Executive Director Tsutomu Masuda addressed the survey results directly, and his comments frame the adoption story with a clear caveat. “This demonstrates that generative AI is being widely and steadily adopted in development environments,” Masuda said, according to AUTOMATON. He also said concerns regarding infringement that arise from generative AI use must be addressed by the game industry as adoption continues, and that a key challenge going forward will be enhancing the quality of human-centered game development while balancing new technological possibilities against the protection of intellectual property rights.
That’s a notably measured statement from the head of an organization whose own survey just reported 85.8% adoption. Masuda isn’t celebrating the number uncritically. He’s pairing it with an explicit acknowledgment that Japan’s copyright framework and the industry’s rights-protection practices haven’t caught up to how widely the tools are already being used. CESA’s preview didn’t disclose a rate of copyright disputes or infringement claims tied to AI-generated assets, so it’s not possible to say how big a problem this currently is in practice. What’s clear is that CESA sees it as the unresolved half of the adoption story.
What Happens Next: The Full Report Lands in December
The figures released on September 17, 2026 are explicitly a preview. CESA’s full 2026 Video Game Industry Report, which typically includes a much wider set of data on market size, employment, and export figures alongside the generative AI findings, is scheduled for release in early December 2026, according to Yomimono’s reporting. That full report should offer a clearer picture of which specific tasks generative AI is used for in 2026, whether the 32% figure for AI-assisted in-house engine development has grown, and whether CESA plans to track infringement incidents or legal disputes as a standalone metric going forward.
Until then, the September preview leaves several open threads: no per-studio breakdown, no cost-savings figure attached to the “reduced development and operational costs” benefit developers cited, and no data on whether generative AI use correlates with headcount changes at CESA member companies. Those gaps matter, because they’re the exact numbers labor advocates, investors, and rival studios abroad will be watching for in December.
5 Predictions for Generative AI in Japanese Game Development
- The full December report will show a further rise, not a plateau. With 63.0% of developers already using generative AI daily, the trajectory from 51% company-level adoption in 2025 to near-universal individual use in 2026 suggests the December figures will land at or above the September preview numbers, not below them.
- Expect a formal CESA position on AI copyright by mid-2027. Masuda’s comments about unresolved infringement concerns read like the setup for a policy statement or industry guideline, not just a passing caveat, especially with member studios exposed to reputational risk from AI controversies like the one Level-5 faced.
- Smaller Japanese studios and contractors will adopt faster in 2027 as tooling gets cheaper. The current 85.8% figure skews toward CESA’s larger member companies. Expect the gap between big publishers and smaller outsourcing studios to narrow as AI tooling costs continue to fall.
- Western publishers will cite CESA’s numbers in their own AI adoption arguments. A concrete, named-source statistic like 85.8% is far easier to cite in a shareholder letter or press statement than vague claims about “industry trends,” and expect US and European studios to reference it when justifying their own AI tooling investments.
- Human-review requirements will tighten rather than loosen. Given that the most common safeguard CESA reported was human verification and correction of AI output, expect that requirement to become more formal, not less, as studios face more scrutiny over AI-generated content appearing in shipped games.
Frequently Asked Questions
What is CESA and why does its survey matter?
CESA, the Computer Entertainment Supplier’s Association, is the Japanese industry body that organizes Tokyo Game Show and represents game hardware and software suppliers, including major publishers like Capcom, Sega, Level-5, and Konami. Its annual Video Game Industry Report is one of the most cited sources of data on the Japanese game industry, which is why a preview figure like 85.8% generative AI adoption carries weight beyond a single studio’s internal statistics.
What percentage of Japanese game developers use generative AI in 2026?
According to CESA’s preview of the 2026 Video Game Industry Report, released September 17, 2026, 85.8% of Japanese game developers use generative AI in their work, based on 1,349 valid survey responses collected between May and August 2026.
How does the 85.8% figure break down?
Of the 85.8% total, 63.0% of developers said they use generative AI on a daily basis, while the remaining 22.8% said they use it occasionally. The two figures combined make up the total usage rate CESA reported.
Which companies were included in the CESA survey?
CESA’s membership includes a large number of prominent Japanese game companies, among them Capcom, Sega, Level-5, and Konami. The survey targeted game developers and CESA member companies broadly, rather than a single studio.
Is Japan’s generative AI adoption rate higher than the global average?
Based on currently available survey data, yes. The Game Developers Conference’s State of the Game Industry survey found 36% of professionals personally use generative AI and 52% said their company uses it somewhere in the organization, both well below CESA’s 85.8% developer-level figure for Japan. South Korea’s Korea Creative Content Agency reported a 41.7% generative AI utilization rate for its game industry in the first half of 2025, also below Japan’s 2026 figure. It’s worth noting these surveys measure adoption in somewhat different ways, so the comparison should be read directionally rather than as a precise ranking.
Does the 85.8% figure represent an increase from last year?
AUTOMATON reported the new figure as an increase from a 2025 CESA figure of roughly 51%. That comparison should come with a caveat: the 2025 figure measured usage among CESA member companies as a whole, while the 2026 figure measures usage among individual developers, so the two numbers aren’t a perfectly matched year-over-year comparison.
What safeguards are Japanese studios using with generative AI?
The most common safeguard developers reported was conducting human verification, correction, and supervision of generative AI output. Many respondents also said they specify or restrict which AI tools are approved for use, and a large share said they avoid using AI-generated output directly without human editing.
When will the complete 2026 CESA report be released?
The full 2026 CESA Video Game Industry Report is scheduled for release in early December 2026. The September 17 figures represent a preview released around Tokyo Game Show 2026, ahead of the complete dataset.
