Journalist |
Editor | Wen Shuqi
An indie developer, five days, and just over a dollar’s worth of API calls—that’s all it took to build a detective game.
That game, called Cold Case Files, walked away with the “Snapdragon Best Tech Breakthrough Award” at the BOOOMJam game jam. In the game, every AI suspect can instantly interpret evidence, hide information, and even fabricate testimony. Stories like this became the hottest talking point at this year’s ChinaJoy.
Industry folks were blown away that AI games have already gotten this good. A solo developer, in almost no time at all, produced something that used to require a small team to pull off.
On August 1st, the line outside the Shanghai New International Expo Centre snaked back over a hundred meters. The 2026 China Digital Entertainment Expo & Conference (ChinaJoy) brought together nearly 900 exhibitors across more than 140,000 square meters of exhibition space, spanning gaming, hardware, chips, robotics, consumer goods—every industry fighting for the attention of young people.
ChinaJoy has now walked through its 22nd year, and this year’s official theme is “Play with AI.” Discussions about AI in gaming, both inside and outside the venue, have hit unprecedented levels of heat. Walking through the show floor, I could feel it: AI has already deeply penetrated the B2B side of the gaming industry, while consumer-facing AI games are standing right on the edge of an explosion.
B-Side Is Fully Saturated, C-Side Awaits Its Native Moment
At ChinaJoy, AI was practically everywhere—from development tools to the robots roaming around, the entire industry chain was flexing its AI muscles.
In Hall N5, Qualcomm booked out an entire venue for the seventh year running. By showcasing the on-device compute and graphics power of Snapdragon PCs and phones, they let players and creators get hands-on with AI-game design and play. Over in Hall E6, I noticed AI companies like Yingmoo Tech, which teamed up with Unity China to announce that, based on their 3D-generation large model, they’d build out engineering capabilities that span from 3D generation to real-time game application.
Deng Qi, an executive director at Shanda Capital under Century Huatong, has been coming to ChinaJoy for over a decade. She told me that AI started seeping into the gaming industry a couple of years back, and in those early days, the focus was mostly on cutting costs and boosting efficiency—like using AI to help write code, generate art assets, or whip up 3D textures.
“But fast-forward to today, whether it’s us or our competitors, AI adoption on the B2B side is already incredibly high.” Right now, Century Huatong’s R&D pipeline has an AI penetration rate of over 85%. On the publishing and operations side, roughly 90% of user-acquisition videos and marketing materials are now generated directly by AI, Deng Qi told me.
That number is almost identical across the industry’s top players. Wang Chuanpeng, VP of Data at 37 Interactive Entertainment’s tech center, revealed that AI-assisted 2D art assets make up over 80% of their output, with a single quarter producing over 500,000 pieces of material and efficiency gains exceeding 80%. AI-deeply-involved ad creatives account for over 70% of their video output.
AI now covers every core gaming role—tech, design, planning, operations, and marketing. Its penetration on the production side has basically become industry consensus.
This year, though, the conversation around AI games has started shifting toward the consumer side, with everyone asking when AI-native gameplay and the so-called “AI Native game” will finally explode. Think back to the mobile era and Angry Birds—it nailed the native touchscreen interaction that truly defined the mobile-game genre. AI games need that exact same “native moment.”
Jian Lvchun, the developer behind Cold Case Files, believes the future of AI games will evolve from a “director’s cut” to “improvisational theater.” Every NPC will have their own motivations, habits, and desires—no more standing around like statues waiting for a player to trigger a quest. Players won’t be passive recipients of a pre-written script anymore; they’ll be a variable inside a living, breathing world.
Deng Qi shared a similar observation. She told me the industry is already seeing some AI gameplay experiments that get teams genuinely excited, like lightweight interactions with AI NPCs where every session feels different and tailors itself to each individual player. But these products are still in the “lightweight” phase—nothing has emerged yet that could define the AI game category as a benchmark.
Over at the Tencent Games booth, I got to try out mini-games that AI had built in just two days and one week. The players around me were genuinely curious about AI games, but their feedback was honest: these games are clearly playable now, yet the flaws are just as obvious—”enemies get stuck after dying,” “after a while you can’t find any enemies and just end up wandering around the map.”
These little broken moments tell a bigger story: AI can crank out games fast, but there’s still a long road to polished, premium quality.
“Everyone’s still in the exploration phase,” Deng Qi told me. “On one hand, the cost of large models hasn’t dropped enough yet. When costs are unpredictable, game companies get cautious about pushing products to users. On the other hand, from the user’s perspective, people are open and willing to try AI games—and that’s a good sign.”
She estimates that mature AI game products hitting the market in a serious way could still be 18 months to two years away. But the pace of technological evolution might outrun our expectations. Anthropic’s latest models can already generate a polished, high-quality game in a single click. With longer context windows, real-time game generation and hyper-personalized interactions might arrive much sooner than anyone thinks.
Tech Gets Democratized, but the Bar Hasn’t Vanished
From demo to full-fledged product, how many hurdles do AI-native games still have to clear?
Talking to game developers at ChinaJoy, I learned the first hurdle is stability—both in AI characters and in the world itself. Chen Mingqiang, CEO of Yuanzai World, pointed out that AI agents struggle with character consistency. In the first few exchanges, the AI sticks to its role, but after longer conversations, you get OOC (out-of-character) moments. World settings drift too—an ancient fantasy setting might slowly morph into modern sci-fi if you’re not careful.
The second hurdle is proving out a viable business model. Bo Honglin, co-founder of China Film Game, pointed out that many demos look flashy but can’t survive contact with reality—the core question is whether they can close the commercial loop. “AI can help you quickly validate a bunch of directions, but at the end of the day, whether users will actually pay and whether you can build a sustainable business—that still comes down to whether the product itself is compelling enough.”
The third hurdle is the sheer unpredictability of the tech frontier. The boundaries are blurry and shifting so fast that product design feels like it’s in a state of perpetual turbulence.
Zhao Tianyi, who heads up manufacturing at TapTap, flagged an interesting comparison: in the past, a 48-hour game jam could only produce a “gameplay slice,” and a 21-day jam could get you a “small demo.” But this year, in the same 48 hours, the quality of AI-assisted projects has completely blown past what used to take three weeks. He also revealed that TapTap has already seen over 4,000 AI-related games listed on its platform in the first half of this year alone.
AI can lower the barrier to entry, but it will never lower the bar for what makes a truly great game. Deng Qi stressed this in our interview: technology will eventually be democratized, but creativity and ideas will always come down to people.
Plenty of industry veterans told me the same thing: the future benchmark for AI games might not come out of a big studio’s massive budget. It’s far more likely to be born from an individual developer—someone who spends almost nothing, has a flash of inspiration, builds a tiny demo in a game jam, and then watches it evolve into a defining piece of an era.
That’s exactly why big companies are now falling over themselves to host AI game “Vibe Coding” challenges—they’re hunting for that raw, individual creative spark. “AI has pulled big studios, mid-size companies, and even solo devs onto the same starting line,” Deng Qi said. “In fact, smaller players without cost pressure or KPI anxiety might just come up with far more interesting ideas than big-company employees.” That said, the big studios’ global publishing experience, localization expertise, and user-acquisition firepower are all accumulated advantages that AI simply can’t replace.
Take a lap around the 2026 ChinaJoy expo halls and you’ll see two things at once: players queuing up in long lines at AI game booths, and developers huddled together debating the technical limits and what future products should look like.
Everyone believes AI will fundamentally change gaming. But right now, nobody can say with any certainty what that “killer app” is actually going to look like.
