Every developer polled in Japan's Online Game Association's (JOGA) 2026 Online Game Market Research Report now uses generative AI, with Google's Gemini leading adoption at 94 percent and Anthropic's Claude at 84 percent.

The figure, drawn from JOGA's annual industry survey and previewed by Famitsu, marks a dramatic escalation from the 2025 Tokyo Game Show survey, when only 50 percent of developers across the wider Japanese industry reported using AI tools.

Microsoft's GitHub Copilot rounds out the top three at 76 percent adoption. According to the report, the leading use cases were data-oriented: "user preference analysis" and "user behavior prediction" — both functions that would have required dedicated analytics teams a decade ago.

What do developers actually use these tools for?

JOGA's data analysis focus is telling. Rather than the asset-generation workflows that have dominated Western AI debates, Japanese online game studios lean heavily on general-purpose large language models for telemetry, segmentation and forecasting tasks that previously demanded large analytics teams.

The pattern fits how online games operate, where retention, live-service tuning and monetisation already run on user behavior prediction and preference analysis. Putting LLMs on those pipelines produces measurable ROI faster than putting them on art production, which still faces consumer resistance.

Capcom has separately confirmed it sees "a certain degree of effectiveness" from generative AI, while emphasising that creative production remains human-led. Sony struck a similar note at a recent annual investor presentation, framing AI as a productivity layer around the studio rather than a replacement for creative staff.

That split — analytical AI in production, human authorship for art and narrative — offers one path through the consumer backlash visible in Western markets. It also explains why Japanese adoption could climb faster: the AI workloads doing the heavy lifting are invisible to players at runtime.

What's worrying Japanese players?

Players polled by JOGA ranked copyright infringement as their top concern around the technology. The second most cited worry: games becoming similar to one another as AI-driven design conventions spread.

Both concerns map onto specific commercial risks. Copyright exposure from training data is the subject of ongoing litigation in the United States and Japan, with rights-holders pursuing damages from model providers and downstream users. Title similarity carries costs around player trust and franchise identity, particularly for live-service titles competing on differentiation.

How does Japan compare to the West?

The contrast with Western studios is sharp. Epic Games is integrating generative AI tooling directly into Unreal Engine 5.8, embedding model access into the engine layer rather than leaving it to individual studios. Against that, developers such as Adhoc Studios, the team behind Dispatch, have publicly refused to use the technology. Larian Studios reversed a pro-AI stance on Divinity concept art after player pressure, opting not to deploy generative AI at all.

Regional sentiment shapes adoption curves. Japanese and South Korean publishers appear to face less organised consumer opposition than their Western counterparts, even as Western developers absorb the brunt of community pushback.

Will the economics hold up?

The economics underpinning JOGA's 100 percent figure may be the most volatile variable. AI vendors face growing pressure to monetise, with Anthropic and OpenAI under investor scrutiny to deliver returns. Cerulli Associates reported in February 2026 that only 38 percent of affluent investors feel comfortable with AI, and Uber has scaled back AI spending after exhausting its token budget as subsidies receded.

Whether Japanese online game studios retain their appetite for Gemini and Claude when unit costs rise is the question worth watching into 2027, when JOGA publishes its next market report and Western publishers release their next financial disclosures.