AI Game Dev News: 85% of Japanese Devs Now Use AI

Welcome to our Weekly AI Game-Dev News Roundup for September 13-19, 2026. Generative AI is quickly becoming normal infrastructure for making games, not a sci-fi experiment. This week shows why beginners should use it as a transparent helper for coding and prototyping, while keeping human judgment in charge of design.

How This Was Verified

This roundup is based on official announcements and reporting; we did not run these tools hands-on. We summarize what publishers and trusted outlets reported and link every claim inline so you can check the original context. All cited URLs returned HTTP 200 on 2026-09-19.

CESA Report: Generative AI Is Now the Norm in Japan

Japan’s biggest signal this week comes from the CESA Video Game Industry Report preview, distributed at Tokyo Game Show 2026 and reported on September 17-18. According to coverage from 80.lv, 85.8% of Japanese game developers now use generative AI at least occasionally, while 63% use it routinely.

That is a big jump from about 51% in the previous 2025 report, as detailed by Automaton Media. CESA is the Computer Entertainment Supplier’s Association, the body that organizes Tokyo Game Show, and its member list includes Capcom, Sega and Konami, according to PC Gamer. The full report publishes in early December 2026.

For beginners, the important nuance is what “using generative AI” actually means. It covers programming assistance, internal ideation, documentation, localization, testing, research, asset processing and prototyping — not necessarily generated art shipped in a finished game, as explained by 80.lv. So do not assume 85.8% of games are now full of AI art.

The 2025 report found visual and image generation was the most-cited use, followed by story and text generation, then programming support, with about 32% of companies also using AI while developing proprietary game engines, according to Automaton Media.

There is also an interesting global contrast. A Google Cloud-commissioned study of 615 developers across the US, South Korea, Norway, Finland and Sweden found 87% using AI agents in their workflows, reported by PC Gamer. That same coverage framed these as dueling surveys — Japanese devs embracing AI while North American ones are still skeptical.

Think of this like version control or a game engine: it is becoming standard workflow support, not a replacement for your design decisions.

Roblox RDC 2026: AI Building Expands Beyond New Zealand

Roblox held its Roblox Developer Conference on September 11, 2026, with a strong push to make AI-assisted creation easier and let people play Roblox games outside Roblox.

The headline for beginners is “Build” — Roblox’s generative AI game-creation feature, which lets users make games from natural language prompts. Build launched in July in New Zealand only, and is now expanding to Serbia and Singapore. It is also gaining desktop access for big-screen creation, a new asset library, and iterative control, and later this year it gets prompt-based scene generation that will extend to Roblox Studio — all per TechCrunch’s RDC coverage. NPCs that can playtest a game are planned for later this year too.

That matters because Roblox creation is already a real economy. Game makers have earned more than $5 billion on the platform since 2013, with $1.7 billion in the last 12 months, according to TechCrunch.

Distribution is opening up too. Games will be publishable as separate apps on mobile, PC and consoles, and by year-end players will be able to join a game in-browser through a link. If you are just starting out, this lowers the barrier to sharing your first playable prototype with a friend who does not have Roblox installed.

Level-5 Apologizes Over AI Footage

This week also brought a clear lesson in transparency. After the September 10, 2026 “LEVEL5 VISION 2026 II” showcase, fans spotted strange visual details in footage connected to Yo-kai Watch 2: Haunted Domain and the Professor Layton and the Curious Village Remake, according to Backyard Drunkard.

CEO Akihiro Hino confirmed on September 12, 2026 that generative AI had been used in event footage and apologized, saying the intent was to make the presentation look more spectacular.

A follow-up statement on September 17 said some flagged details were actually intentional simplification or human production errors. That same update stated clearly that AI was NOT used directly in the game content — Professor Layton and the New World of Steam specifically did not use AI, according to Backyard Drunkard’s breakdown.

For new developers, the takeaway is simple: players can tell when promo material does not match the real game, and they expect honesty about it.

Also This Week: Garry’s Mod Goes 64-Bit

Not everything this week was about generative AI. Garry’s Mod received its September 2026 Update on September 16, 2026, according to the official Facepunch news post.

The biggest technical change is 64-bit Windows builds for the game client, Hammer, and dedicated servers, which means less risk of running out of 32-bit address space and crashing on heavy servers. If you have ever had a modded sandbox crash when you spawned too much stuff, that address-space limit is often the culprit.

The update also adds Sandbox improvements, including a new Hacked Manhack NPC, Spawnmenu sub-categories, and an overhauled player model selector, plus Insurgency from 2014 added to the mountable games list, per Facepunch. For tinkering and learning systems design, Garry’s Mod remains a wonderful, low-pressure playground.

What This Means for Solo Devs

You do not need to copy big studios, but you can borrow their safest workflows. Start with low-risk AI uses like code help, docs, localization, prototyping, research, and testing assistance — the same categories highlighted in the CESA coverage from 80.lv. Use AI to explain an error message, draft a quest description you will rewrite, or generate placeholder dialogue while you test your combat loop.

If you are choosing your first engine, browse our beginner-friendly framework guides before you add any AI tools, so you understand what the AI is helping you with. Then share a small playable game early, even if the art is simple gray boxes.

Most importantly, disclose AI use and never pass generated footage off as shipped content. Label prototypes as prototypes, keep AI mock-ups out of your trailer unless you clearly say what they are, and remember the Level-5 lesson: the apology was about presentation footage, not about using AI to brainstorm. Transparency builds trust, and trust is everything when you are solo.

FAQ

Is generative AI now required to make games in Japan?

No. The CESA numbers show AI assistance is common for tasks like coding help, docs, and prototyping, not that finished Japanese games are AI-generated. Learn core design and programming first, then add AI carefully for specific bottlenecks where you can review the output.

Can I build a Roblox game just by typing a prompt?

You can start that way with Roblox Build, which creates experiences from natural language prompts and now supports desktop editing and iteration. For beginners, treat prompts as a prototype starter, then learn Roblox Studio basics to fix, polish, and understand how your game actually works.

Should I disclose it if I use AI in my indie game?

Yes, be transparent. Level-5’s apology shows players notice and care when promotional footage uses AI without disclosure. Label AI-assisted placeholders, keep records of your workflow, and never present AI mock-ups as final gameplay to avoid losing trust with your small community.