Generative AI has stopped being a side experiment for Japan’s game studios and become the default way work gets done. A preview of the 2026 CESA Video Game Industry Report, shared ahead of Tokyo Game Show 2026, shows that 85.8% of Japanese game developers now use generative AI tools in some part of the development process, as confirmed as of September 2026, up from 51% in September 2025. That is a 34.8-percentage-point jump in twelve months, according to the Computer Entertainment Supplier’s Association (CESA), Japan’s main game industry body. The figures were reported by AUTOMATON, and picked up in translation and summary form by outlets including Gamekyo, IG News, and 3DNews.
The number lands at a pointed moment. Tokyo Game Show has spent the last few years positioning itself as a showcase for both traditional console blockbusters and AI-assisted production pipelines, and this year’s preview data gives that framing a hard statistic to point to. For an industry that spent 2023 and 2024 debating whether generative tools belonged in professional pipelines at all, an 85.8% adoption figure reads less like early experimentation and more like a completed transition.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Headline Number: 85.8% and Rising Fast
CESA’s 2026 survey breaks the 85.8% adoption figure into two bands. About 63.0% of Japanese game developers say they use generative AI on a daily basis, while another 22.8% use it occasionally. Combined, that’s the 85.8% total figure that outlets have led with. A year prior, CESA’s 2025 survey found that 51% of Japanese game companies were using AI or generative AI in game development, based on responses from 54 companies surveyed. The jump between those two data points, from roughly half the industry to nearly nine in ten developers, is the core news here.
Framing varies slightly by outlet, but the substance doesn’t. AUTOMATON’s English-language coverage headlined the story around the “increase from last year’s 51%.” French-language aggregator Gamekyo emphasized efficiency gains and shorter production cycles as the story behind the number. Russian outlet 3DNews tied the developer-adoption figure to a parallel claim about the share of newly released Japanese games built with generative AI assistance, framing the growth as a 34-point swing in a single year. Korean business coverage tied to Tokyo Game Show 2026 used the 85.8% figure to argue that Japan’s studios are now treating AI tooling as standard production infrastructure rather than a novelty.
What makes the shift notable isn’t just the size of the jump. It’s the speed. Most technology adoption curves inside conservative, deadline-driven industries like game production move in single-digit percentage steps year over year. A move from 51% to 85.8% in twelve months suggests something closer to a coordinated, industry-wide decision rather than gradual, studio-by-studio experimentation.
Who Publishes the CESA Report and Why It Matters
CESA, the Computer Entertainment Supplier’s Association, is Japan’s primary trade body for the video game industry. Its membership includes major publishers and developers such as Capcom, Sega, Square Enix, Konami, Bandai Namco, SNK, and Level-5, among others. CESA’s annual Video Game Industry Report is one of the few structured, association-backed data sources tracking how Japanese studios actually build games, rather than relying on anecdote or social media sentiment. The organization also runs Tokyo Game Show jointly with Nikkei Business Publications, which is why the AI adoption figures were previewed in connection with this year’s event.
The methodology CESA uses for its AI-adoption tracking targets both individual game developers and CESA member companies directly, a design that captures both large studio behavior and the experience of individual working developers. The 2025 precursor survey, which produced the 51% baseline, drew on responses from 54 Japanese game companies. CESA has since confirmed that its 2026 findings rest on a considerably larger sample: 1,349 valid developer responses collected as of September 2026, a scale that lends the 85.8% figure more statistical weight than the prior year’s company-level snapshot, extended to a wider developer and company base ahead of the full report’s release at Tokyo Game Show.
That association backing matters for credibility. Individual studio claims about AI usage are easy to dismiss as marketing. A trade association aggregating data across dozens of member companies, including some of the best-known names in console and mobile gaming, is a different kind of signal, and it’s part of why the figure spread so quickly through gaming press in Japan, Korea, France, and Russia within days of the preview.
Where Generative AI Is Actually Being Used
The CESA data and related industry commentary point to three dominant categories of generative AI use inside Japanese game studios: visual asset generation, text and narrative generation, and programming support. Image generation for characters, environments, and concept art was the single most common use case identified in CESA’s prior-year reporting, followed by story and dialogue text generation, then code-assistance tools for programming tasks. That ordering appears to hold into 2026, even as the overall adoption number has grown sharply. Individual studio moves back up the trend: Capcom confirmed in October 2026 that it had integrated generative AI directly into its proprietary RE ENGINE to optimize internal tools and shorten development time, while Koei Tecmo launched RyzaChat: AI in August 2026, a consumer-facing generative text-and-voice companion app built on the same category of tooling CESA’s survey tracks.
A related and more granular survey covering Japanese online game companies, published by Famitsu in July 2026 and reported separately by AUTOMATON, found 100% generative AI adoption within that specific segment. In that survey, Google’s Gemini was used by 94% of respondents, Anthropic’s Claude by 84%, and GitHub Copilot by 76%, making them the three most common tools in that sub-industry. Notably, the strongest reported use cases in the online game segment weren’t art or code generation at all, but user preference analysis and user behavior prediction, an analytics-driven application of generative AI that gets far less attention in general coverage of AI in games.
Beyond the headline categories, CESA’s broader reporting has also flagged generative AI as a tool developers are exploring for in-house engine development. In prior-year data, 32% of surveyed companies said they were looking at AI assistance for building or extending an in-house game engine, a use case that sits closer to tooling infrastructure than to content generation and suggests studios see applications well beyond art and dialogue.
| Use Case | Reported Adoption Signal | Primary Beneficiary |
|---|---|---|
| Visual/art asset generation | Most common use case cited in CESA reporting | Art and concept teams |
| Story and dialogue text generation | Second most common use case | Narrative and writing teams |
| Programming/code assistance | Third most common use case; Copilot cited at 76% in online-game segment | Engineering teams |
| User preference and behavior analysis | Top use case within online game companies (100% AI adoption segment) | Live-ops and analytics teams |
| In-house engine development support | 32% of companies exploring this application (prior-year CESA data) | Tools and infrastructure teams |
Daily Use vs. Occasional Use: What the Split Really Shows
The distinction between the 63.0% of developers using generative AI daily and the additional 22.8% using it occasionally is arguably more informative than the combined 85.8% headline. A near two-thirds daily-use rate indicates that generative AI has become embedded in routine production workflows rather than being reserved for occasional brainstorming or one-off asset generation. That’s a meaningfully different claim than “developers have tried AI tools,” which is closer to what adoption surveys in other regions and industries typically measure.
Denfaminicogamer’s original Japanese-language coverage of the CESA figures leaned specifically on this daily-use framing, describing the more than 60% daily-use rate as evidence that generative AI is now embedded in development workflows rather than treated as an experimental add-on. That framing, more than the topline 85.8%, is what separates this report from earlier, more tentative adoption surveys that simply asked whether a company had used an AI tool at any point.
It’s worth being precise about what the daily-use figure does and doesn’t claim. It does not mean 63% of developers use AI for the majority of their workday, nor does it specify which roles within a studio are driving that daily usage. CESA’s preview data, as reported, does not break the daily-use figure down by discipline (art, code, design, QA), so any claim beyond “the categories are art, text, and code, in that order” would be reading more into the numbers than the association has published.
Why Adoption Jumped 34 Points in One Year
CESA’s reporting and the coverage built around it point to three recurring reasons developers and studios give for the shift: improved operational efficiency and productivity, shorter development cycles, and reduced development and operational costs. Those three factors appear consistently across French, Korean, and Japanese-language coverage of the same underlying data, suggesting they reflect the actual survey responses rather than press interpretation.
Game development has always been resource-intensive relative to output, with large teams spending months on asset iteration, localization, dialogue writing, and playtesting cycles that don’t scale linearly with headcount. Generative tools that can accelerate concept art iteration, draft dialogue variants, or assist with boilerplate code offer a direct lever against those cost and time pressures, particularly for mid-sized studios that can’t staff every discipline at AAA scale. That economic logic, more than novelty or hype, is likely what pushed adoption from roughly half the industry to nearly nine in ten developers in a single reporting cycle, a dynamic also tracked by industry outlets like GamesIndustry.biz covering production-cost pressure across the wider global market.
There’s also a peer-pressure dynamic worth naming plainly. Once a critical mass of competing studios adopt AI-assisted workflows and see production-cycle gains, staying out becomes a competitive liability rather than a cautious choice. CESA’s own membership includes companies competing directly for the same talent, publishing slots, and player attention, which creates strong incentive for laggards to catch up quickly once the leaders demonstrate results.
Copyright Concerns Haven’t Slowed the Numbers
Not everything about the report is celebratory. Coverage tied to the CESA figures notes that copyright questions around AI-generated content remain unresolved as a legal and ethical matter within the Japanese games industry, even as adoption climbs. That tension, unresolved legal questions alongside near-universal practical adoption, is unusual, and it mirrors debates happening in game development markets outside Japan, including ongoing disputes in North America and Europe over AI training data and asset ownership.
Discussion around the CESA figures has also surfaced internal studio-level debate about the reliability and originality of AI-generated art and text, along with concerns about the effect on entry-level roles that have traditionally served as training ground for junior artists and writers. Those concerns echo a controversy that played out publicly earlier in 2026 around a high-profile Japanese sequel announcement, where fan and industry reaction to AI-assisted art direction became a flashpoint for exactly this debate, illustrating that adoption at the studio level and acceptance at the player level are not moving at the same pace.
What the data suggests, though, is that studios are largely choosing to manage that risk rather than avoid generative AI altogether. Sega offers a concrete illustration of that posture: the publisher formed an internal generative-AI committee in 2025 to test the technology across code, images, and motion work, but said in October 2026 that it had deliberately restricted generative AI’s production use to publishing, operations, and planning tasks rather than core creative asset work. An 85.8% adoption rate, reached in a single year despite an acknowledged and unresolved copyright question, indicates that Japanese studios have judged the productivity upside worth the legal ambiguity, at least for now. Whether that calculation holds if a major copyright dispute produces an unfavorable court ruling in Japan or a key export market is one of the more consequential open questions hanging over the report.
How Japan Compares to Global AI Adoption in Game Development
Coverage of the CESA figures consistently frames Japan’s 85.8% developer-level adoption rate as notably higher than global estimates for generative AI use in game development, which in most regions still cluster at or below the 50% mark. That comparison should be read carefully: CESA’s own 2025 baseline of 51% was itself already roughly in line with what would be considered “early majority” adoption internationally. The story in 2026 isn’t that Japan is ahead of the curve in absolute terms, it’s that Japan moved from roughly matching the global pace to substantially outpacing it within a single year.
That divergence raises a fair question about what’s structurally different in Japan’s game industry that would produce a faster adoption curve than comparable markets. Candidate explanations include CESA’s coordinated, association-driven approach to surveying and (implicitly) normalizing AI tool usage across its membership, a games industry culture with historically tight production timelines and cost discipline, and a wave of mid-sized studios facing exactly the kind of resource pressure that makes AI-assisted workflows most attractive. None of those explanations are confirmed by the CESA data directly, but they’re consistent with the productivity, cost, and cycle-time reasons developers themselves cited.
| Metric | 2025 (CESA Baseline) | 2026 (CESA Preview) |
|---|---|---|
| Overall generative AI adoption among developers | 51% | 85.8% |
| Daily generative AI use | Not separately broken out | 63.0% |
| Occasional generative AI use | Not separately broken out | 22.8% |
| Companies surveyed | 54 Japanese game companies | 1,349 valid developer responses (CESA, September 2026) |
| Companies exploring AI for in-house engine work | 32% | Not yet broken out in preview data |
Tokyo Game Show 2026: The Backdrop for the Announcement
The CESA adoption figures were shared as part of the industry-facing material tied to Tokyo Game Show 2026, one of the largest game industry events in Asia and a joint production between CESA and Nikkei Business Publications. That puts it in the same late-year cluster of major industry showcases as Europe’s Gamescom, which sold out its 2026 edition with record attendance. Coverage connected to the show, which runs in the Chiba prefecture exhibition halls tracked on the official Tokyo Game Show event site, describes it as a major stage this year for AI-driven production and “subculture” genre games, with strong interest from Korean and other Asian developers attending or exhibiting. Tokyo Game Show 2026 opened its doors on September 17, 2026, and the show floor included multiple generative-AI development vendors demonstrating production tooling alongside traditional publisher booths. One of the clearest examples of that AI-driven production theme arrived the same day the show opened: Nexon Games unveiled Pareidolia on September 17, 2026, a title built around AI voice-responsive characters, giving the CESA adoption figures an immediate, playable showcase rather than just a survey statistic.
Tokyo Game Show has increasingly become a venue where CESA uses its convening power to shape industry narrative, not just showcase upcoming titles. Publishing the AI adoption preview ahead of the full report, timed to the show, ensures the data gets maximum visibility among the exhibitors, publishers, and international press who attend specifically to gauge where the Japanese industry is heading next. That timing strategy is itself a signal: CESA appears to want the 85.8% figure to become the dominant industry talking point of this year’s show, ahead of individual game reveals.
What This Means for Global Studios and Publishers
For studios and publishers outside Japan, the CESA figures function as a competitive benchmark whether or not they were intended as one. If a market with Japan’s production quality standards and famously meticulous QA culture has moved to 85.8% generative AI adoption in a single year, studios elsewhere weighing whether to formalize AI tool usage now have a hard comparison point, a trend also being tracked by outlets like TechRadar covering AI tooling across the broader tech industry, rather than an abstract trend to point to internally.
That comparison cuts both ways commercially. Publishers negotiating co-development or localization deals with Japanese studios should expect AI-assisted asset and text pipelines to be the default working assumption going forward, not an exception to flag in a contract. At the same time, the unresolved copyright question CESA itself has flagged means international partners will likely need clearer contractual language around the provenance of AI-assisted assets before those deals close, particularly for markets like the US and EU where AI-training-data litigation is already active.
Historical Context: From Cautious Experimentation to Near-Universal Use
Generative AI’s path into professional Japanese game development has been faster than most enterprise software adoption curves typically move. Two years ago, generative AI in games was largely discussed in Japan, as elsewhere, in terms of concept art experimentation and early code-completion tools, with meaningful internal skepticism about production-ready reliability. CESA’s own 2025 survey capturing 51% adoption already represented a notable jump from the prior informal estimates that put usage in the low double digits. The move to 85.8% represents the point where generative AI stopped being a differentiator between forward-leaning and conservative studios and became close to universal baseline infrastructure, similar to how version control or middleware engines are treated today.
That kind of compressed adoption timeline, less than two full survey cycles to go from early-majority to near-universal, is unusual even by software industry standards, and it puts Japan’s game development sector ahead of the adoption curves reported in most enterprise software categories tracked by other 2025-2026 industry surveys, where generative AI tool usage more commonly grows in the range of 10-20 percentage points annually rather than the 34-point jump CESA reported here.
Competitive Landscape: How AI Tool Providers Are Positioned
The tool-level data, published by Famitsu in July 2026 from its survey of the parallel online-game-company segment, gives a useful, if narrower, window into which AI providers are actually winning adoption inside Japanese studios. Google’s Gemini led that segment at 94% usage, ahead of Anthropic’s Claude at 84% and GitHub Copilot at 76%. That ordering is notable because it puts a general-purpose multimodal assistant (Gemini) ahead of a code-specialist tool (Copilot), reinforcing the broader CESA finding that art and content generation, not just programming assistance, is driving the bulk of adoption. It’s also a reminder that model choice matters for accuracy-sensitive tasks like localization and dialogue QA, where hallucination rates vary meaningfully between leading chatbot models.
It’s worth flagging that this tool-level breakdown comes from the narrower online-game-company survey, not the full CESA developer-wide 85.8% figure, so it shouldn’t be read as a precise map of tool share across the entire Japanese game industry. Still, directionally, it suggests that studios adopting generative AI broadly are relying on a mix of large general-purpose model providers alongside specialized coding assistants, rather than consolidating around a single vendor. Investor interest in that tooling category has followed the same adoption curve: General Intuition raised $220 million in a September 2026 funding round that valued the game-video AI startup at $6.2 billion, a sign that capital is flowing into the specialized AI infrastructure Japanese studios are increasingly built around, not just the general-purpose chat assistants topping Famitsu’s usage table. For studios weighing which image models to standardize on for concept art and asset pipelines, a broader look at the current field of AI image generation tools is a useful starting point, and studios running multi-tool pipelines can also learn from how Western teams are building multi-model AI image pipelines to avoid single-vendor lock-in.
Predictions: Where Japan’s AI-in-Games Story Goes Next
Based on the trajectory CESA’s data shows and the broader dynamics around it, a few outcomes look likely over the next 12 to 18 months.
- Adoption is likely to approach a ceiling near or above 90% within the next survey cycle, since the remaining non-adopters are increasingly likely to be very small teams or studios with specific ideological objections rather than companies unaware of the tools.
- Expect CESA or a comparable body to publish more granular, role-by-role adoption data in future reports, breaking down daily use by discipline (art, code, writing, QA, analytics) rather than reporting a single aggregate daily-use figure.
- Copyright and provenance disputes tied to AI-generated game assets are likely to become a more visible flashpoint in Japan over the next year, following the pattern already seen around individual high-profile titles that faced fan backlash over AI-assisted art.
- International publishers and co-development partners will increasingly formalize AI-disclosure and provenance clauses in contracts with Japanese studios, given how close adoption now is to universal.
- Other major game-development markets, particularly South Korea and parts of Southeast Asia, are likely to see accelerated adoption reporting in 2027 as studios compete for the same efficiency gains Japanese developers are now citing.
What CESA’s Report Doesn’t Tell Us Yet
It’s worth being direct about the limits of the preview data. CESA has not yet published the full 2026 Video Game Industry Report, only a preview of top-line adoption figures ahead of Tokyo Game Show. CESA has at least clarified the scale of its sample, 1,349 valid developer responses as of September 2026, but several other important details remain unconfirmed as of this writing: a discipline-by-discipline breakdown of the 63% daily-use figure, quality or output measurements tied to AI-assisted work, and any data on job impact or headcount changes tied to the shift. Readers and industry analysts should treat the 85.8% figure as directionally strong and well-sourced through CESA, while waiting for the full report for granular detail.
The gap between preview and full report also explains some of the inconsistency across international coverage. Outlets reporting quickly off the preview data, translating from Japanese-language sources, or summarizing secondary coverage have understandably emphasized different angles, from the raw percentage jump to the copyright controversy to the tool-level breakdown from the separate online-game-company survey. The full CESA report, once published in connection with Tokyo Game Show 2026, should resolve most of those gaps.
Frequently Asked Questions
What percentage of Japanese game developers use generative AI in 2026?
According to a preview of the 2026 CESA Video Game Industry Report, 85.8% of Japanese game developers now use generative AI in some part of their work, up from 51% in 2025.
What is CESA?
CESA, the Computer Entertainment Supplier’s Association, is Japan’s leading video game industry trade body. Its membership includes major publishers such as Capcom, Sega, Square Enix, Konami, Bandai Namco, SNK, and Level-5. CESA also co-organizes Tokyo Game Show.
How many developers use generative AI daily versus occasionally?
Of the 85.8% total, 63.0% report using generative AI on a daily basis, while 22.8% use it occasionally, according to the CESA preview data.
What do Japanese game developers use generative AI for most?
The most common applications, based on CESA’s reporting, are visual/art asset generation, story and dialogue text generation, and programming assistance, in that order. A separate survey of online game companies also found strong use in user preference analysis and behavior prediction.
Which AI tools are Japanese game companies actually using?
A survey of Japanese online game companies found Google’s Gemini used by 94% of respondents, Anthropic’s Claude by 84%, and GitHub Copilot by 76%. This data comes from a narrower segment of the industry, not the full CESA developer-wide survey.
Is this adoption rate higher than other countries?
Coverage of the CESA figures frames Japan’s 85.8% rate as notably higher than global estimates for generative AI use in game development, which in most regions are still reported at or below roughly 50%.
Are there concerns about AI-generated content in Japanese games?
Yes. Coverage tied to the CESA report notes that copyright questions around AI-generated content remain unresolved, and there has been public debate in Japan over the reliability and originality of AI-assisted art and writing, echoing controversies seen around specific high-profile game announcements earlier in 2026.
When will the full 2026 CESA report be published?
The figures reported here are from a preview shared ahead of Tokyo Game Show 2026. CESA has not yet published the complete Video Game Industry Report with full methodology and detailed breakdowns as of this writing.
