Why so many game developers don't want to use generative AI

Generative AI, though pitched as a revolution, remains taboo. Game developers like Stardew Valley's Eric Barone and Subnautica 2 studio Unknown Worlds openly avoid or malign it. The makers of trophy-adorned RPGs like Kingdom Come: Deliverance 2 and Baldur's Gate 3 faced severe backlash over whiffs of gen AI. Nvidia DLSS 5's AI-powered filter stoked fierce rejection among devs and players.

Other game developers and, especially, executives do see value in AI. Bethesda boss Todd Howard called it a potentially useful "tool," just not for generation. But the CEOs of EA and Ubisoft have championed AI as the tech of the future, with the latter investing heavily in "player-facing generative AI" specifically. 2025 hit Arc Raiders famously used loads of AI voices (and later replaced many with people), and Summer Game Fest 2026 has seen myriad AI-aided games.

Getting game developers to talk about gen AI is not always easy. In the process of interviewing over 30 devs on the subject, many people refused to speak to me or requested anonymity. I wanted a thorough examination of the tech, and kept core questions to just three: How do you feel about gen AI in games and in game development? Do you want to use it? And how do you want it to be treated in the games industry?

Perhaps pro-AI developers didn't want to talk to me or I just didn't run into any during my survey, because I heard an overwhelmingly negative assessment of generative AI's origins, capabilities, and risks. By the end, I'd heard dozens of developers make a case against using gen AI at all.

What's wrong with gen AI?

With their enormous physical server stacks, AI data centers guzzle water for cooling and incur massive power costs – with electricity draw for just one center equaling tens of thousands of homes – while the infrastructure required puts pressure on local communities and environments. These centers have also disrupted hardware markets and supply chains, leaving consumers facing price hikes as AI projects make key components scarce. RAM prices, for instance, have more than tripled in under two years. Meanwhile, AI-generated assets frequently rely on references which were scraped without creator permission. And many of the biggest AI providers, from OpenAI's ChatGPT to xAI's Grok, face legal action over what their AI tools produce, all while companies using these tools are grilled over how implementing AI often means firing people – sometimes after using them to train AI.

Iron Lung and Dusk creator David Szymanski, who is "not categorically against AI as a whole technology," believes it needs to "clean up its act" before it can begin to be taken seriously. Like many devs, Szymanski is unwilling to "hand wave all the ethical concerns about plagiarism, environmental impact, and job security" tied to gen AI.

When you generate words, images, or sounds with many of today's biggest AI tools, you can't know exactly what, or how many, references were used for that asset, and that may include a wealth of copyrighted work. David Gaider, former Dragon Age narrative lead turned Summerfall Studios co-founder, focuses on the moral issue. Artists did not consent to having "their data pillaged," he says, and he's not moved by the argument that AI "won't work as well" if it can't scrape literally everything. The counterargument is obvious: if rampant theft is integral to the tech as-is, perhaps it should not be allowed to exist as-is.

The legal hazard is also enormous. Games have been pulled from sale over lapses in music copyright; serious regulation or litigation of AI-generated assets could shatter entire studios. To many, the tech isn't worth the copyright landmine. Marvel Rivals executive producer Danny Koo says this is partly why the hit hero shooter doesn't use gen AI at all. "I think it is better for us to very confidently say all the assets in Marvel Rivals are created by everyone on the team without any assistance from the outside, so it's not poisoned," he explains.

There's another major risk to using gen AI in game development and beyond. When companies boast about replacing staff with AI, first of all, they may be lying, as Take-Two CEO Strauss Zelnick argues, and simply using AI as a smokescreen for cost-cutting. But when AI does truly stand in for people, it's often put on junior-level tasks. Wage-free labor might sound great to shareholders in the short-term, but multiple developers warn that eliminating junior positions jeopardizes the long-term health of the industry.

Gaider worries, "How are we going to train up the next generation of devs if we eliminate every entry-level task?" Rami Ismail, veteran dev and consultant and once half of studio Vlambeer, sees "the collapse of the junior pipelines" brewing. And this is happening at a time when the games industry is already struggling with brain drain at the top after cataclysmic layoffs, with key devs exiting studios, or the industry altogether, and taking invaluable knowledge with them. Further cannibalization of junior roles by gen AI could shut out would-be talents, squeezing the industry's workforce from both ends and making a career in games even tougher to navigate.

Why aren't devs using gen AI?

If you can get past these concerns, devs say gen AI still cannot do what people can. Several who have tried it say the technology fails key variety and quality tests, and because it cannot create – it can only derive and imitate – it's unavoidably behind in a creative industry where it's already hard to stand out. We've heard the case for using AI for conceptual or non-final assets, but that brings concerns about the foundation you're laying for yourself with that approach.

Gaider reasons that AI's inability to iterate consistently, capture intent, and establish replicable processes makes it impractical for something as complex as game development. "It would be frustrating as hell" to make games if you couldn't trace how you got certain results, he says. That's important for making more than one thing in isolation, and especially if you want to get better at making things. But because it is purely focused on output, gen AI use spirals: the more you use it, the more you have to use it, because by using it you avoid learning the skills to make things without it.

"It wouldn't be so bad if generative AI was seen more as an assistant," Gaider finds, but often he sees people trying to "clean up" after AI is left to do the "important work" instead, which takes time and leads to "mediocre" outcomes. He adds: "It's not ready for prime time. There's just a lot of executives who really, really want it to be."

New Blood head Dave Oshry says "we were fine for decades without it and we don't need it now," bluntly calling the tech "garbage." Koo reiterates that preserving the identity of Marvel Rivals and the passion of the dev team is priority. Ismail won't use it, he explains, because "I like my job" and won't offload it to a machine.

Similarly, Pocketpair and Palworld comms lead John Buckley finds gen AI unnecessary because "we have a lot of artists" who like "doing stuff themselves." Buckley has fielded gen AI accusations throughout Palworld's life; he says Pocketpair doesn't use it and there's no discussion to be had at all. He reckons using AI as a search tool or coding assist is "a very different conversation," but on gen AI, he stresses, the case is shut: "Gamers don't want it."

"This gen AI side of gaming feels kind of similar to the early crypto stuff," says Buckley. "It feels very intrusive. It feels like everyone who is super gung-ho about it isn't from the industry. They're, dare I say, outsiders looking to get rich quick. Where I think it's very different in this case is that gaming lives and dies by the consumers, the players, and the players say they don't want it."

Szymanski interrogates a common sales pitch. "Right now the siren's song of AI is that it's really fast and really easy," he observes. "There's definitely a place for fast and easy, but there's also a place for slow and difficult." Szymanski believes creativity emerges when ideas meet problems. "If you have a copying algorithm solve all those problems for you, what exactly sets your finished work apart from the things the algorithm copied?"

Whether or not it's being put to good use, gen AI in games is on the rise. A lead developer at a large studio who wished to remain anonymous believes all devs, even those opposed to AI, should examine what it is capable of, arguing "we have no choice because it's in your face" and "it's starting to impact how we work every day." Their "personal feeling is also mixed," but they hope that a better understanding of the tech will lead to "more right decisions" made about it.

Nick Herman, co-founder of Dispatch maker AdHoc Studio, says his team is "not exploring" gen AI as an option, and "there's no version where we're ever gonna just write a script for AI". Despite this rejection, he wants to be open-minded about new technology. He remembers the early Photoshop days: some people said if you used the software, "you're a hack, that's not real art," but now "if you don't know Photoshop, you're not gonna get a job."

"I personally don't have some hard-coded opinion that's so strong I'm not allowing that to change over time." Yet he sees no reason to use AI creatively, and can't shake a familiar feeling: "I definitely do have an initial reaction that this isn't how art should be made."

The effects of gen AI

When gen AI is used in games, it routinely fails to meet expectations. It's not necessarily that AI has no use cases whatsoever, but many devs told me it's being oversold and under-regulated, and they resent seeing it forced into roles it's not suited for, often despite resistance from people in those roles. Many devs have been told by higher-ups that AI will unlock their creativity, but a lot of those devs are expected to find the proof of that themselves. Yet heavy-handed integration can instead make AI an obstacle.

Sam Barlow, a former Silent Hill writer now known for Her Story and Immortality, calls the current uses of gen AI "inefficient and broken." He proves Gaider's point, mentioning a colleague who was asked to "clean up" AI-generated dialogue for a lower rate on a shorter timeline. "But the actual work was much harder than writing everything fresh because they had to keep trying to figure out what the original intent was or fix things that made no sense at all," he says.

Gen AI has been held up by some as an antidote to the swelling production costs and timelines in the industry, particularly for AAA games. Strange Scaffold boss Xalavier Nelson Jr., a prolific dev who's worked on over 100 games, says gen AI won't solve these time or money problems because it doesn't address their real cause, and if anything, could enable more excess.

"Generative AI is a can button, not a should button, and every game you have loved is built off of shoulds," he argues. "For all the people who say it will let us finally make our big games faster and cheaper as the costs go exponentially up, the unspoken decision that has occurred is that rather than ask whether the size of the game is feasible, and whether it provides the thing that is most meaningful for a player, they have decided that the sheer scope is the god of their development process and it must be fulfilled at all costs."

Bruce Straley is a Naughty Dog veteran best known for games like Uncharted 4 and The Last of Us, and he's now heading up Coven of the Chicken Foot as an indie. On gen AI, he admits, "I'm scared, is how I feel. I'm scared for what it means for everything that I've learned about game development. I'm scared for the old ways that I've thought about how to build a team, what I've learned about skill and craft." But, he maintains, "you still need people to have taste, and you still need people to understand craft to be able to get to an end result that is worth anything," pushing back on the "magic bullet" narrative.

As I weigh the 32 interviews transcribed for this article against the endorsements showered on AI, much of what I see reads like conflict between people who want to do the work and those who just want to wish things into being. On top of the moral, environmental, and legal concerns, many devs do not understand the desire to outsource the processes that drew them to this field in the first place. "I always felt game development is one of those jobs you only do when you really want to do it," says Ismail, as "you're very unlikely to get rich in an industry that lays off people for making superhits".

Developers told me they could not find a productive creative application for AI. That could be because the proposed efficiency gains have not materialized, the impact on team morale can be devastating, player goodwill can sour in an instant, or AI-generated work is reliably inferior. Even if all of this improved and gen AI became much more convincing and sustainable, there are still concerns of painting yourself into a corner by becoming reliant on volatile technology. There are measurable consequences and creative implications attached to generative AI, and people clearly care about how their art is made.

"We cough up a chunk of our soul": 32 game devs, from Doom's John Romero to Helldivers 2 and Palworld leads, explain what people get wrong about games.