> ## Content Index > Fetch the complete content index at: https://blog.kyleelliott.net/llms.txt > Use this file to discover other available public pages before exploring further. # 85.8% of Japanese Game Developers Use Generative AI. What Does That Number Actually Tell Us? - URL: https://blog.kyleelliott.net/85-8-of-japanese-game-developers-use-generative-ai-what-does-that-number-actually-tell-us/ - Published: 2026-10-02T02:15:16.000Z - Updated: 2026-10-02T02:15:16.000Z - Description: CESA says 85.8% of Japanese game developers use generative AI. I'm a player, not a dev, so I look at what that number really measures, why player research disagrees with itself, and where I draw my line. - Author: Kyle Elliott - Tags: gaming, ai I play a lot of games. JRPGs, TRPGs, shooters, platformers, action games, and a good number of indie titles in those genres. I'm not a game developer and I've never made a game. I'll also be honest about something else. I have no idea whether any game I've played used generative AI. That's why a number from this month caught my eye. At Tokyo Game Show 2026, CESA (the industry group that runs the show) released a preview of its Game Industry Report. The headline stat: 85.8% of the Japanese game developers it surveyed use generative AI in their work. That's a big number. It's also an easy one to over-read. So here's what I think it says, what it doesn't, and where I land as someone on the player side of the screen. ## What the 85.8% Actually Is The figure comes from a survey with 1,349 valid responses. Of those, 63.0% said they use generative AI routinely and 22.8% said they use it occasionally. Add them up and you get 85.8%. A few things to keep in mind. **It's self-reported and counted per person.** It doesn't say how much of anyone's work involves AI or what the AI touches. "We use this tool" can mean anything from running everything on it to trying it once. Nearly a quarter of that headline number is the "occasionally" group. **"Game developer" is a wide net.** The survey covered producers, directors, engineers, artists, sound creators, designers, QA testers, and executives, plus some educators and students. A mix of roles is a good thing. It also means the number describes a broad group, not one studio's pipeline. **The jump from last year is unverified.** A lot of headlines said usage rose from 51%. The coverage I read disagrees on whether that comparison holds. One report says CESA's free 2025 release has no generative AI usage percentages. Another describes the 51% as a company-level figure from a different survey base. I can't confirm the two numbers are comparable, so I'm not leaning on the "jump." **I'm working from coverage of the preview.** I haven't read CESA's own report. The full version is due in early December 2026\. I'll revisit this when it lands. CESA also surveyed its member companies, though only 48 of 220 responded. Efficiency and productivity was the most expected benefit, named by 38 of them. The most common management measure was human verification, correction, and review. Copyright and intellectual property ranked as the top challenge. ## Players Don't Agree With Each Other Research on how players feel about generative AI is all over the place. That matters before drawing conclusions from the developer number. A preprint on arXiv looked at 508,192 English-language Steam reviews. Games that disclosed generative AI had lower recommendation rates and more negative sentiment than games using procedural generation. Players often described the AI use as low developer investment. Two caveats. The comparison is against procedural generation, not games with no AI at all. And reviews can't separate the effect of AI from the quality of the game. The paper also notes more than 10,000 Steam games carried AI disclosure labels as of February 2026. Quantic Foundry surveyed 1,799 gamers between October and December 2025\. 85% had a below-neutral attitude toward generative AI in games, and 63% picked the most negative option. Opposition was strongest for creative uses like art, music, and narrative. It was weaker for things like dynamic difficulty adjustment. It was an optional survey taken after viewing a gamer profile, so the sample is self-selected. MIDiA Research found something different. In a survey of 6,300 gamers at the end of 2024, most were neutral, and it rarely affected purchases. Its analysis suggested the loudest negativity came from developers, influencers, and journalists, not everyday players. All three can be true at once. The samples differ, the timing is a year apart, and Quantic itself noted that attitudes seemed to have worsened over the previous year. I also couldn't find solid research specific to indie games. That's a real gap, since so many of us seek them out. ## My Line Quick reminder of where I'm coming from. I don't have a game to point to. My own hands-on use of generative AI is limited to code generation for applications. So this is a player's view, not an insider's. My line is simple. If it's good and it doesn't stand out, I don't see the issue. Some people like disclosure labels. I'm indifferent to them. That fits one reading of the Steam study, where players described AI use as low investment. Maybe what bugs people is perceived effort and quality more than the tool. But the study doesn't prove that. Quantic's results suggest some players object to the use itself, especially in art, music, and writing. My preference is one data point, not a trend. ## Why I Think It's Here to Stay I land somewhere in between. People might not like it, but I think it's here to stay. As the tools get better, I expect to see more of this kind of output. That's a prediction, not a finding. One detail from the coverage backs it up. The most common safeguard companies reported was human review, and many cited practices like restricting which tools are used and avoiding direct use of generated output. My inference, not CESA's, is that a lot of that 85.8% may sit upstream of what ships. Think prototyping, tooling, and code. If so, usage can keep climbing while players see very little of it. That would also make disclosure labels less useful over time. If nearly everyone uses AI somewhere in the pipeline, a label tells you less about what you're actually playing. That's speculation on my part, but it's the part I find most interesting. ## What I'd Like to See Next CESA's full report in December, with a breakdown by task and a clear comparison to earlier years. That would tell us more than any single percentage. ## Sources - [AUTOMATON West: Over 85% of Japanese game developers use generative AI, 2026 CESA survey shows](https://automaton-media.com/en/news/over-85-of-japanese-game-developers-use-generative-ai-in-game-development-2026-cesa-survey-shows-an-increase-from-last-years-51/?ref=blog.kyleelliott.net) - [Inven Global: 85.8% of Japanese game developers use generative AI, CESA's first survey](https://www.invenglobal.com/articles/26191/858-of-japanese-game-developers-use-generative-ai-for-work-cesas-first-survey?ref=blog.kyleelliott.net) - [Digital Citizen: CESA says 85.8% of surveyed Japanese game developers use generative AI](https://www.digitalcitizen.life/cesa-survey-japanese-game-developers-generative-ai/?ref=blog.kyleelliott.net) - [Player Perceptions of Generative AI in Games: A Steam Review Analysis (arXiv preprint)](https://arxiv.org/pdf/2608.11539?ref=blog.kyleelliott.net) - [Quantic Foundry: Gamers and generative AI](https://quanticfoundry.com/2025/12/18/gen-ai/?ref=blog.kyleelliott.net) - [MIDiA Research: What do gamers really think about generative AI in games?](https://www.midiaresearch.com/blog/new-research-what-do-gamers-really-think-about-generative-ai-in-games?ref=blog.kyleelliott.net)