A JOGA market report says every surveyed Japanese online game company is using generative AI, with user preference analysis and behavior prediction leading the use cases.
Image: foro3d.com
Every surveyed Japanese online game company reported using generative AI
A new Japanese game developers AI survey has put a hard number on a trend the industry usually discusses in softer terms: according to Famitsu’s preview of the JOGA Online Game Market Research Report 2026, 100% of surveyed Japanese online game companies said they use generative AI tools.
The finding comes from the Japan Online Game Association’s 22nd annual market research report, published by JOGA and Kadokawa ASCII Laboratories on July 10, according to Automaton West. JOGA has run its online game market survey since 2004, which gives this year’s AI section extra weight. This is not a one-off social media poll or a platform holder’s marketing deck. It is an industry report focused on Japan’s domestic online game market.
The tension is in the scope. The headline number sounds like a verdict on the whole Japan game industry AI landscape, but JOGA’s report is narrower. VGC notes that JOGA defines the field as domestic Japanese games “played via the internet regardless of device,” while also stressing that the report does not cover console or PC games outside that online-game framing, nor offline mobile games that do not connect users online. In practical terms, the survey speaks most directly to companies operating connected games, where user data, live events, retention pressure, and service tuning sit close to the center of development.
That distinction does not deflate the result. It explains it. Online games already run like living machines, with developers reading player behavior, adjusting economies, testing content cadence, and trying to keep communities from drifting away. Generative AI game development in that environment is less about a robot drawing the hero’s cape and more about a studio trying to read the battlefield faster.
Gemini, Claude, and GitHub Copilot lead the reported toolset
Famitsu’s coverage, as reported by VGC, Eurogamer, and Automaton West, says Google Gemini was the most widely used generative AI tool among surveyed companies, with 94% reporting use. Anthropic’s Claude followed at 84%, while GitHub Copilot was reported at 76%.
Those names matter because they complicate the popular image of AI in video game development. The most visible arguments around generative AI often focus on concept art, voice, music, scripts, and copyright-sensitive asset generation. The JOGA results, at least as described in the available coverage, point to broad-purpose productivity and analysis tools rather than a single specialized art generator dominating the production floor.
GitHub Copilot’s presence at 76% fits the programming side of development, where code completion, scripting assistance, and repetitive implementation tasks are natural places for AI tooling to enter. Gemini and Claude suggest a wider spread of uses, from summarizing information and drafting internal materials to querying patterns in player data, though the cited report preview identifies the most common delegated work more specifically: user preference analysis and user behavior prediction.
Automaton West adds a useful year-over-year contrast. In the previous year’s JOGA survey, according to Automaton’s summary of Famitsu, “content planning” was one of the top cited uses alongside user analysis, and ChatGPT was the most-used tool at 59% utilization. This year’s reported lead for Gemini at 94%, plus the shift toward user analysis and behavior prediction, suggests the center of gravity has moved from brainstorming assistance toward operational decision-making for online games. That is an interpretation of the reported figures, not a confirmed cause stated by JOGA in the provided source material.
The biggest use case is player behavior, not replacement artwork
The most important detail in the JOGA report preview is not the 100% figure by itself. It is what those companies are reportedly asking AI to do. According to Famitsu’s coverage cited by VGC, Eurogamer, and Automaton West, the tasks companies most commonly delegated to generative AI were analyzing user preferences and predicting user behavior.
For online games, that is the command room. Preference analysis can inform which characters, modes, events, difficulty curves, cosmetics, rewards, or monetization beats appear to be landing with players. Behavior prediction points toward retention modeling, churn risk, event timing, matchmaking patterns, and the delicate rhythm of when to challenge players, when to reward them, and when to let the game breathe.
From an action and adventure design lens, that matters because pacing is not only a campaign concern anymore. In a connected game, pacing is also a service problem. A boss encounter may be tuned around completion rates. An event may be scheduled around when players return after work or school. A reward track may be adjusted because the data says a grind is becoming a wall. AI-assisted analysis does not automatically make those decisions good, fair, or player-friendly, but it can make the feedback loop faster.
The survey coverage does not say that Japanese online game developers are handing final creative authority to generative systems. It also does not say that AI-authored assets are the dominant use case in this sample. The available reporting instead points to a quieter and potentially more pervasive role: AI as a layer between player data and production choices. That is less cinematic than a machine producing concept art, but it may affect the shape of live games more directly.
Players are worried about copyright and sameness
JOGA’s report did not only ask companies about AI. According to the coverage cited by VGC, Eurogamer, and Automaton West, it also surveyed players of online games about generative AI use in the games they play.
The concerns are familiar but pointed. Respondents reportedly raised fears that games infringing copyright could become more likely, and that future games may start to resemble one another because of generative AI use. Those worries line up with the two pressure points that have followed AI tools across creative industries: provenance and homogenization.
The copyright concern is easiest to understand. If a game uses generated art, text, sound, or other creative material, players may wonder what source material shaped the output, whether the studio has rights to use it, and whether human creators were displaced or copied without consent. The JOGA coverage provided here does not identify specific infringing games or allege that surveyed companies are using infringing outputs. It reports player concern about the risk.
The sameness concern is more slippery and more relevant to design. If multiple studios lean on the same general-purpose tools, prompts, models, optimization goals, and behavior predictions, they could be nudged toward similar content structures. In online games, that might mean familiar event schedules, similar reward loops, comparable character archetypes, or safe narrative beats designed to satisfy predicted preferences. The risk is not that every game instantly becomes identical. It is that optimization can sand down odd edges, and odd edges are often where memorable games get their rhythm.
Japan’s broader AI adoption was already climbing before this report
The JOGA result lands in a market where AI adoption had already been rising. Eurogamer points to the 2025 annual Tokyo Game Show survey, where 50% of surveyed developers across a broader range of game companies said they were using generative AI. Automaton West cites a Computer Entertainment Supplier’s Association survey from September that found 51% of Japanese game companies were using AI in some capacity, with common uses including generation of visual assets and images, followed by story and text generation.
Those broader surveys are not directly interchangeable with JOGA’s online-game sample. They cover different populations and, based on the available reports, different definitions of AI use. Still, the contrast is striking. A roughly half-and-half industry-wide picture becomes universal adoption inside the polled online-game company group.
Company comments also show how Japanese publishers have been trying to frame AI adoption. Eurogamer reports that Capcom previously said it had seen a “certain degree of effectiveness” from generative AI while emphasizing that creative parts of development would remain human-led. Eurogamer also notes that Sony expressed a similar human-led position in a recent annual investor presentation. Automaton West separately reported Capcom engineers describing AI as a way to reduce routine tasks rather than match creators’ sensibility.
That public language is careful. It presents AI as a production assistant, a tool for routine labor, and a way to support human teams. JOGA’s reported use cases fit that framing in one sense, since user analysis and behavior prediction are operational tasks. They also raise a harder question: when a connected game’s roadmap is increasingly shaped by predictive systems, where exactly does creative leadership begin and end?
The number is clear, but several questions remain unanswered
The confirmed claim from the cited reporting is specific: Famitsu’s preview of the JOGA Online Game Market Research Report 2026 says 100% of surveyed Japanese online game companies use generative AI tools. The most-used tools were reported as Gemini at 94%, Claude at 84%, and GitHub Copilot at 76%. The most common delegated tasks were user preference analysis and user behavior prediction. Players surveyed by the report raised concerns about copyright infringement and games becoming similar.
What is not answered in the provided material is equally important. The excerpts do not provide the survey’s sample size, the exact wording of every AI-use question, the split by company size, the degree of use, or whether companies are using these tools in shipped production, internal planning, analytics workflows, coding assistance, or all of the above. A studio using Copilot for internal scripts and a studio using AI-generated assets in a live product would both sit under a broad “uses generative AI” umbrella, but those are very different stories for players, workers, and rights holders.
There is also a disclosure gap. VGC cites Google Cloud’s global director for games, Jack Buser, who told Mobilegamer.biz that practically every major game development studio now uses AI, while not all are comfortable disclosing it because the subject is divisive. Buser claimed Google’s survey around Gamescom found roughly nine out of 10 developers saying they used AI, while other surveys have landed closer to 40% to 50%, a difference he attributed to willingness to say so. That is an executive claim from a company selling AI tools, so it should be read with that interest in mind, but it matches the broader pattern: reported use is rising, disclosure remains uneven.
For players, the practical guidance is to read future AI claims with attention to scope. “AI was used” can mean code assistance, analytics, localization drafts, internal ideation, asset generation, testing support, or live-service prediction. The JOGA survey’s most revealing contribution is that, among Japan’s surveyed online game companies, AI appears to be embedded in the systems that study players as much as in the tools that help developers build. That is where the next debate over online game developers AI tools is likely to move: away from whether AI exists in the pipeline, and toward how much influence it has over the games people return to every night.