The AI “silver bullet” in gaming is already biting us, and I’m done staying quiet

The exact moment I realised the AI conversation had gone completely off the rails

I knew the AI discourse in games was broken the day I read an investor say he was “shocked and sad” that developers were demonising generative AI.

Shocked and sad. Not about layoffs. Not about people’s art being scraped without consent. Not about the environmental cost of training yet another billion-parameter model so some exec can brag on an earnings call. No, he was shocked and sad that we – the people who actually make and play games – weren’t applauding loudly enough for his new money printer.

I’ve been gaming since the Quake and PS1 era, and I’ve worked around game tech long enough to remember when “AI” meant your Halo Elite actually flanking you instead of walking into a wall. So when I see venture capital guys clutching pearls because devs are wary of generative AI, I don’t see visionaries being unfairly maligned. I see people who don’t know – or don’t care – how games are actually built telling the rest of us we’re ungrateful for not biting the silver bullet they’re trying to ram into the chamber.

And the irony is, games have quietly been one of the most advanced AI playgrounds on the planet for decades. The current narrative – that AI is some new arrival that stubborn devs are blocking – is nonsense. Worse than that, it’s dangerous nonsense, because it gives the people selling the tech all the power, and leaves the people who have to live with the fallout looking like Luddites.

I’m not anti-AI. I’m anti-bullshit. And right now, the generative-AI gold rush around games is about 20% interesting tech and 80% bullshit. If the industry doesn’t wrestle back control of the story, we’re going to wake up in a few years wondering why we have bigger games, cheaper content, and somehow… worse experiences and fewer jobs.

Newsflash: AI has been in games longer than half the people hyping it

One of the most surreal parts of this whole debate is listening to platform holders and investors talk about AI in games like it’s a “shiny new feature”. Meanwhile, anyone who was around for the late ’90s and early 2000s is sitting there going: mate, we’ve been doing this.

Think back:

• Quake III Arena bots that actually strafed, dodged, and felt like players
• Half-Life soldiers coordinating and flushing you out with grenades
• F.E.A.R. enemies dynamically reacting, flanking, and retreating
• The Sims juggling dozens of needs and behaviours under the hood
• Halo 2 – backed by Xbox Live’s TrueSkill matchmaking to quietly sort you into lobbies where you belong

That was “AI” long before some guy in a Patagonia vest started talking about large language models on stage.

Then came motion matching – first used in big games and now baked into engines – where animation systems use machine learning-style selection to stitch player moves into something that feels fluid and natural. We’ve had ML quietly running servers to catch cheaters, detect toxicity in chat, and power smarter matchmaking. Massive mobile studios have been using player analytics and ML-driven profiling for years to optimise live games and content pipelines.

Some studios have QA bots hammering new builds all night, ML-based tools to profile GPU performance across platforms, and internal systems that sift through absurd amounts of assets in seconds. None of this is “coming soon”. It’s here. It works. It’s boring – in the best possible way – because it’s embedded into workflows, not sprayed across a trailer as a marketing bullet point.

So when I hear “AI will transform how games are made”, I always think: it already did. The part you’re late to is not the tech – it’s the conversation. And that’s exactly why we shouldn’t let the people who just showed up dictate how the next wave gets used.

The generative gold rush: a hammer looking for a game-shaped nail

Since GPT‑4 dropped, every earnings call sounds like a parody. “We’re exploring how generative AI can revolutionise our content pipeline.” “We’re excited about infinite, bespoke player stories.” Translation: “Our investors are asking where the AI bullet point is and we need a slide that sounds impressive.”

You can see the appeal. Generative AI promises the holy trinity every exec dreams about:

• Make games faster
• Make games cheaper
• Make games bigger

Suddenly you’re sold visions of NPCs that chat forever, dynamic quests that never run out, and worlds so dense you’ll never see the same thing twice. All it takes is a sprinkle of prompts and some GPU clusters heavier than small countries.

Meanwhile, the ugly parts get hand-waved away: the fact that models hallucinate, regurgitate copyrighted work, and cost a fortune to train and run; the lawsuits piling up around training data; the energy draw so huge that even AI CEOs are out there saying things like, “raising humans uses more energy than training AIs” and then somehow wondering why people hear that and go, “wow, you sound like you literally see us as a wasteful meat-based benchmark.”

Add on top the hardware crunch – DRAM and GPUs being hoarded for data centres, delaying consumer hardware and pushing up prices – and you start to see the shape of the problem. This isn’t free. Somebody pays for this experiment, whether it’s players buying more expensive consoles, devs being laid off because “tools can do more with less”, or entire studios being steered toward “AI-driven” designs because that’s where the investment winds are blowing.

And here’s the cruel twist: even when the tech is genuinely clever – the “world model” stuff that auto-generates game environments from videos, or models that can sketch out interactive levels – it’s miles away from being usable as-is in a real pipeline. It’s fragile, inconsistent, and expensive. But because someone’s sunk billions into it, suddenly we’re all supposed to pretend this is the future of game development.

The myth of the magic content cannon (and the biting silver bullet game)

Let’s tackle the biggest lie head-on: generative AI is not a magical content cannon. You don’t point a model at your design doc, press a button, and get a finished level / questline / cutscene. If you’ve ever actually tried to use these tools for anything serious, you know how quickly the fantasy collapses.

The outputs are inconsistent. You spend ages prompting and re-prompting, then editing, then checking for copyright issues, then trying to get the thing to stop making up lore or inventing bugs that don’t exist. For every decent idea that drops out, you’re babysitting a system that doesn’t understand context, pacing, or what makes a game feel right.

Execs keep selling generative AI as a silver bullet, but it’s more like a biting silver bullet: bite down hard, and it cracks your teeth. The “biting silver bullet: game” is simple – you swallow the hype now, then discover down the line that you didn’t buy a tool, you bought a dependency. On a third-party vendor. On a black-box model. On a pricing structure that can and will change when they decide they want to be profitable.

Execs keep selling generative AI as a silver bullet, but it’s more like a biting silver bullet: bite down hard, and it cracks your teeth. The “biting silver bullet: game” is simple – you swallow the hype now, then discover down the line that you didn’t buy a tool, you bought a dependency. On a third-party vendor. On a black-box model. On a pricing structure that can and will change when they decide they want to be profitable.

And that’s before we even touch the ethics. Using scraped art and writing without consent and calling it “training data” isn’t clever, it’s parasitic. Putting people out of work with what is, in a lot of cases, a worse version of their craft, isn’t “innovation”, it’s cost-cutting dressed up as disruption. When half the industry is on edge, wondering if they’re next in line to be “augmented”, something has gone very wrong.

Here’s the part nobody in the PR deck wants to say aloud: if you’re using generative AI to replace an artist or writer directly, instead of to support them, you’re not being “future-facing”. You’re just rolling the dice that your audience won’t care about the drop in soul, texture, and intent for the sake of shaving a few months off production. That’s not vision. That’s gambling.

Where AI actually makes games better (and doesn’t screw people over)

Here’s the twist: I like a lot of AI in games. I want more of it. I just want it in places where it makes sense, not because an investor wants a talking point for their next interview.

There are already examples of generative and ML tools that actually pull their weight:

• Search and organisation: multimodal asset search that lets devs find “that one destroyed sci‑fi crate with yellow stripes” in seconds across thousands of assets. That’s huge time saved, no jobs lost.
• Production plumbing: tools that merge duplicate bug tickets, clean up reports, or auto-tag issues. Boring? Good. That’s exactly where AI belongs – gnawing through the grunt work.
• QA and testing bots: ML-powered bots stress-testing builds, hammering servers, or running through levels to catch crashes. They don’t replace human QA, they give them more signal and less noise.
• Upscaling and optimisation: image reconstruction and ML-based upscaling like DLSS-ish tech to make ports viable or reduce texture sizes without torpedoing quality. Humans still decide the look; the machine helps them get there on limited hardware.
• Moderation and safety: smarter toxicity filters, cheat detection that doesn’t just blanket-ban legit players. If AI can keep more real humans safe and playing, good.

Even on the more “creative” side, there are interesting experiments – teams building on-device, narrowly trained models to power specific NPC behaviours, or to generate branching text in tightly controlled spaces with clear boundaries and oversight. Those can be cool. They live or die on design, not on marketing slogans.

The common thread? These tools respect two core principles:

• They augment humans instead of pretending to replace them.
• They’re built on clear, constrained goals instead of “what if AI did everything?” vibes.

That’s the line for me. AI that lets animators focus on polish instead of wrestling with blend trees? I’ll champion that all day. AI that exists because some exec wants to fire half the writing team and still ship a 60‑hour RPG? Miss me with that.

Reclaiming the narrative: we can’t leave AI storytelling to people who don’t play

Right now, the loudest voices shaping the AI-in-games story are:

• Big tech CEOs desperate to justify mind‑melting AI spend
• Investors who see games as content fodder for their models
• PR teams who think “AI-powered” sounds futuristic, even if they can’t explain what it actually does

You know whose voices almost never get top billing? The actual devs trying to ship something that won’t get roasted for AI slop. The artists whose portfolios just got hoovered into a training set. The players who keep saying, over and over, “we don’t want this if it makes your game worse.”

At GDC surveys, more than half of developers say they view generative AI negatively or with heavy scepticism. That’s not because they don’t “get it”. It’s because they do get it – they see where it fits, and where it’s a liability. They’re living with the reality that every time an exec says “AI will help us be more efficient”, what people hear is “your job is on a countdown.”

So what does “reclaiming the narrative” actually look like?

• Studios set their own AI red lines. No training on non‑consensual data. No replacing entire disciplines with models. AI is for workflow support and new experiences, not pure headcount reduction.
• Trade bodies grow a spine. Stop parroting vendor slides and start publishing concrete guidelines: what’s acceptable, what’s exploitative, what must be disclosed.
• Transparent communication with players. If your game uses gen AI, say where and how. Don’t bury it in a footnote. Let people vote with their wallets informed, not tricked.
• Education over fear-mongering. Teach devs – especially juniors – what these tools are good at and where they’re trash. Fear thrives on vagueness; clarity is how you stop people panicking.

Because here’s the thing: anti-AI sentiment among players is not some irrational knee-jerk. It’s the sane reaction to a landscape where they see layoffs, low-effort AI art, soulless writing, and execs bragging about “efficiencies” as if that’s a feature. If we keep letting AI vendors drive the story, we’re going to deserve the backlash that follows.

Where I draw the line as a player and a critic

After hundreds of hours poured into games that feel hand-crafted – the weird little jank in a Shenmue sidequest, the specific flavour of a FromSoftware NPC line, the way a good fighting game character expresses personality through frame data and animations – I’m not trading that in for mass-produced, statistically plausible slurry.

If you tell me your game uses AI to help your animators iterate faster, to find bugs sooner, to keep cheaters out, or to pull off a port that otherwise wouldn’t run on weaker hardware? I’m listening. I want those wins. That’s AI working in service of the craft.

If you tell me your game uses AI to write most of its dialogue, to crank out NPCs at scale, or to replace actors with synthetic voices so you don’t have to pay performance rates? I’m out. I don’t care how “impressive” the tech is in a vacuum. I’m not rewarding you for hollowing out the parts of games that feel human so you can hit a budget target.

And I’m done pretending those are two sides of the same debate. One is pragmatic, ethical adoption. The other is what happens when you let vendor decks and investor pressure define your roadmap.

The games industry has always been at its best when it quietly absorbs technology and bends it to its will. Physics, online networking, shaders, procedural generation – all of that was scary and new once. We tamed it by putting designers, artists, and players at the centre of the conversation, not investors and keynote speakers.

We need to do the same with AI – especially generative AI – before the decisions get made for us. Because they will. Through policy. Through consolidation. Through “AI-first” mandates handed down from people who measure success in “engagement hours” instead of memories.

I care about this because I don’t want to be sitting here in ten years, looking at a shelf full of technically stunning, procedurally padded, emotionally empty games and thinking: we did this to ourselves. We let people who don’t actually play shape what playing looks like.

AI is not the enemy. But the story being told about AI in games right now? That absolutely is. And unless developers and players start pushing back – loudly, specifically, and unapologetically – the biting silver bullet everyone keeps selling is going to hit something we can’t patch later.