I’ve spent the last year bouncing between two brain states. One is “wow, this AI stuff is magic,” and the other is “please unplug it before it turns everything into a beige soup.” Depending on the week, the previews, and the pitches in my inbox, AI looks like gaming’s next great buff or the debuff we never recover from.
What’s Happening Today?
If AI in gaming were an RPG loot table, most of what’s dropped so far isn’t mythical swords or cursed relics but utility gear. Useful. Practical. Not going to make your friends gasp when you equip it, but it’ll keep you alive in the dungeon. In other words, AI is perfect for grunt work.
Ubisoft’s Ghostwriter is basically that NPC companion who carries all the junk so you can focus on the good loot. It cranks out placeholder barks for guards and side characters, freeing up human writers to spend their brainpower on the story beats that actually matter. Then there’s the new breed of chatty NPCs, powered by NVIDIA’s Ace and partners like Inworld. Think of them as co-op buddies who never log off. Basically, AI companions who know their players. I’ve seen demos where they can drive the jeep, point out loot, and even hold a conversation that doesn’t sound like it was ripped straight from a bad 2006 RPG. Still early days, but you can see the bones of something genuinely fun here. In the background, QA testing is quietly leveling up. Razer’s AI QA Copilot and EA’s internal tools are running tests, logging bugs, and helping prioritize fixes so players never have to download a “Day One Patch” the size of a small moon. On the art side, procedural assists are making texture upscaling, animation smoothing, and asset rigging faster. It’s not glamorous, but every minute saved here is a minute that can go into the boss fight animations, environmental storytelling, or that weird little side quest someone on the team is obsessed with. And finally, game balance and player modeling. The invisible hand tweaking difficulty curves and predicting when you’re about to rage quit. Not the sexy stuff you put in a trailer, but the reason you keep saying, “Okay, just one more run.” That’s the current loot pool. Tools that, in the right hands, can free up human creativity instead of replacing it.
That’s the hero case. AI as a productivity buff. You still need designers with taste and writers with teeth. You just ship fewer bugs and spend fewer afternoons writing boring scripts. I’m bullish on a few other lanes as well. First, simulation and systemic AI. Using models to stress-test economies, predict exploits, and tune live-ops feels like a clean win. Accessibility benefits with dynamic subtitles, on-the-Fly difficulty guidance, and smarter hint systems could be huge. And tools for Solo and small teams might be the most transformative. An artist using an AI assistant for fiddly UVs, a designer using world-models to block a level before sculpting it by hand, a two-person studio running nightly AI playtests to catch softlocks while they sleep. Notice that none of that replaces taste.
What Devs Think
When I talk to studios, the mood is pragmatic, bordering on wary. The GDC State of the Industry survey this year showed more teams using generative tools inside companies, but fewer developers saying AI is a positive for games (13 percent, down from 21 percent the year prior). That whiplash tells you something. Wired’s reporting captured the vibe neatly. Rising negativity around generative AI is tied to quality concerns and job fears. Another industry survey noted a jump in respondents who think gen-AI is hurting the industry, pointing to IP theft, energy burn, and mid outputs that flood stores with bloat. I’ve heard all of this privately, too, often from devs quietly using AI for grunt work while loudly reminding me they’ll never use it for story.
And I truly hope so. When I close the game, do I remember a person on the other end of the design? A boss animation that says “someone obsessed over this wind-up,” a line read that lands because the writer knew exactly when to shut up, a map that folds back on itself with a smile. Conversational NPCs could pass that test if they’re framed right, like tight scopes, strong character bibles, and latency that doesn’t step on timing. AI models could help designers sketch faster. But faster is not better. If AI saves time and that time is reinvested into taste, pacing, and feel, players will cheer. If AI saves time and that time funds eleven more battle passes, players will notice that too.
Here’s What I Don’t Want
A few days ago, I watched a demo for Genie 3. I was very impressed. Not necessarily with the demo because it’s still a lot of work, but with the concept, and how far Google pushed it. For those who don’t know, Genie 3 from Google DeepMind lets you turn a prompt into an explorable 3D world you can run around in at 720p/24fps. Sounds like the fever dream of every kid. You type a sentence, it spits out a world, and you’re walking through it in real time. It even remembers what you’ve done for a bit. It’s genuinely wild to poke at.
You can summon weather, nudge objects, and improvise a Tiny sandbox on the fly. But it currently remembers only a sliver of history, struggles with consistent designer intent, and costs real compute to run.
Could Genie (and cousins like it) train AI agents or accelerate blockouts? Absolutely. Will it “design” a stealth mission with the patience of Dishonored’s clockwork mansion or the inevitability of a Dark Souls shortcut that loops back like a magic trick? Not on its own. And I hope it never truly will because it will rob us of so many things that are good in gaming.
If something like Genie 3 becomes a real game development tool, I think AI will most likely turn into a villain in gaming for the majority of gamers. And somehow, when thinking about AI in gaming, I always circle back to that majestic bastard Hideo Kojima. I remember sitting during the pandemic in 2019, roped to a tower of boxes, enjoying every moment of Death Stranding. Hideo Kojima turned isolation into a mechanic and then built a tender social layer out of other people’s footprints. Death Stranding’s main premise shouldn’t work in a game, but it worked because you felt small in a beautiful world, connected to ghosts you never met. That’s the point, and yes, you can quote the man himself on that feeling. An LLM can approximate the vibe, but the leap from “cargo delivery sim” to “meditation on grief and connection” is a human one.
The Villain I Fear the Most
Let’s be realistic. The stock market is the biggest enemy of gaming. Spreadsheets. The real danger is executives confusing content with craft and seeing AI as a line item to cut heads. That future gives us stores filled with infinite filler and NPCs who can talk about everything except the one thing the scene needs. Players will bounce. Devs will burn out. And the games we remember will get rarer. The risk is real, and surveys confirm it. Adoption is up because time is money. Enthusiasm is down because the heart is harder to budget. If studios treat AI like a junior teammate who never sleeps, they’ll be tempted to hand it the pen. AI isn’t creative right now. It’s clever. It’s fast. It’s helpful. It’s also derivative by design. It predicts, but it doesn’t originate. It can scaffold a mood, but it can’t decide what a game means.
So yes, AI can be a hero if we keep it in a support role. It turns villain when it’s abused as a replacement for the messy, human parts of making games, taste, intent, the point of the thing.
If we flood the market with generated filler, we’ll drown out the voices willing to risk a strange, specific idea, like Hideo Kojima does. Which is why I keep coming back to that lonely walk in Death Stranding. No model “comes up” with that. A person does, then a team builds the scaffolding so you can feel it. The best future for AI in games is the one where it gives those people and teams more time to be weird, specific, and unreasonably ambitious. If we want a hero, we don’t need smarter robots. We need braver humans.
