Back in 2024, not long after laying off five percent of the company’s workforce, EA CEO Andrew Wilson got on a stage to wax poetic about all the ways generative AI would help take the gaming industry to the next level. Machine learning tools, he said, sticking to a script that may as well have been spat out by an LLM, would give developers an “exponentially bigger canvas upon which to create, and richer colors so they might paint more brilliant worlds.”
He’s hardly the only executive who’s promised his customers the world in recent years, though how exactly AI is going to impact game development remains as ambiguous now as it was the day we were first introduced to ChatGPT. While some, from EA and Ubisoft to Remedy and Larian, prepare to ride the wave, others are putting up walls and turrets. Hooded Horse’s Tim Bender has called the technology “cancerous” and added a No AI Assets clause to the company’s publishing contracts. Meanwhile, a survey by Game Developer found that 36% of industry professionals are using GenAI in their day-to-day work, but that half of them think it’s making games worse, not better.
Right now, it’s possible we’re headed for a future in which many triple-A titles will be incredibly large but feel lacking in substance, akin to the early days of No Man’s Sky’s procedurally generated, endless universe. Wilson himself suggested as much when he mentioned that College Football 25 – with its 150 stadiums and 11,000 athlete avatars – could not have been made without help from GenAI. Our ever-inflating expectations for larger maps and more realistic-looking graphics have pushed development cycles for franchises like Grand Theft Auto and The Elder Scrolls into the double digits, and AI seems like an obvious and cost-effective way to bring those numbers back down again while delivering gargantuan amounts of content.
But while AI could probably program testicle physics for horses, it can’t create a character as deeply human as Arthur Morgan. Consequently, the more you try to automate workflow, the less human (and therefore engaging) your games become. For this reason, I believe we should expect another, equally influential shift to sweep across the industry, one that will push some developers into the opposite direction. Instead of giant, hollow worlds with glossy AI finishes, I suspect we’ll see more games that are smaller in scope, tighter in design, and a little rougher around the edges. In short, we’ll see games that go where AI cannot follow.
To form an impression of how AI will impact games tomorrow, we can start by looking at how the technology is being used today. According to that Game Developer survey, AI use among industry professionals varies by level of seniority, with upper management using AI tools more frequently than those below them. Across the board, people are more reliant on AI for researching and brainstorming than actual asset generation. In other words, the more complex and creative a task, the less AI is used.
Another Game Developer survey, conducted in 2025, found that nearly half of industry professionals fear extensive use of AI would decrease the quality of their games. Brandon Sheffield, founder of Necrosoft Games and director of Demonschool, argued that overreliance on AI tools will cause games to become more generic in style and design, as these tools tend to slavishly replicate their training data without putting an original spin on them. Hidden Door CEO Hilary Mason, also interviewed for the survey, called AI “aspirationally mid,” adding that a technology which lacks both vision and ambition cannot produce anything other than mediocre slop.
People from various creative industries agree that AI tools remain just that: tools, used to take care of mindless busy work so that employees are free to focus on the tasks that they alone can do. This includes writing emails or cleaning up code, but also generating reference material and altering the color, lighting, or perspective of concept art and other types of assets. Take one look at social media, though, and you’ll see that AI’s role in that “busy work” is just as contested.
We can also learn a thing or two about AI’s impact on game design by looking at the impact of previous technological revolutions, like the invention of the camera. The camera did not, as many painters in the early 19th century feared, destroy the art of painting, but merely forced it to evolve. Now that there existed a device which could, in an instant, capture the likeness of anyone and anything, and do so with greater precision than even the most talented artists, there was no good reason to keep on painting realistically. Rather than compete with technology, art went in a direction where technology could not follow: towards abstraction and subjectivity, towards Impressionism, Expressionism, Fauvism, Cubism, Surrealism, and other movements that represented not the world as it is, but how we as individuals experience it.
Tomorrow’s best and brightest game designers, like yesterday’s most celebrated painters, will search for the limits of what AI can achieve and boldly set up shop outside that boundary. Right now, AI is predictable, regurgitating information in consistent, clearly recognizable patterns, so games made by humans will become more unpredictable and surprising. Genre titles – easily mimicable on account of their fixed rules and tropes – will give way to games that evade straightforward categorization. Just as we no longer use labels and describe almost all visual art as “postmodern,” so too might gaming arrive at a point where we can no longer distinguish shooters from Soulslikes or platformers from puzzlers. In many cases, the lines dividing these terms are already much blurrier than they were in previous decades.
AI cannot have personal experiences, so games will likely become more and more grounded in personal experience. Unlike, say, Pixar’s recent move away from autobiographical storytelling in favor of “mass appeal” – a decision which, judging by the box office performance of the studio’s latest film, Elio, isn’t exactly paying off – developers pushing for authenticity will want to draw from ideas they cannot get from LLMs. Think Ryan and Amy Green’s That Dragon, Cancer, about their child’s battle with terminal illness, Adam Robinson-Yu’s A Short Hike, inspired by memories of past hiking trips, or even Cory Barlog basing Kratos’ relationship with Atreus on his own struggles entering fatherhood.
Chances are, developers will also become increasingly keen on exploring the darker, messier, more confusing aspects of human nature and existence. The side that AI – being non-human – cannot comprehend. ChatGPT has the emotional complexity of your most obnoxious LinkedIn connection, and could never – as journalist Ioan Marc Jones points out in The Bookseller’s article about AI’s expected impact on writing – start a story like Albert Camus begins The Stranger: “Maman died today. Or yesterday maybe, I don’t know.” Think, here, of Papers, Please, which lets players explore the psychological conundrums of working as a border patrol officer in a semi-dystopian society, or Hellblade: Senua's Sacrifice’s nuanced portrayal of psychosis.
Finally, AI-generated content often looks polished and perfect to the point of actually being kind of ugly and uncanny, so tomorrow’s games might embrace flaws, imperfections, and other signs of hands-on tinkering that let the humanity behind them shine through. Indeed, if AI won’t speed up triple-A development through automation, it may well do so by convincing AI resisting perfectionists to let go of some of their more compulsive, time-consuming tendencies. Red Dead Redemption 2 didn’t need horse testicle physics to reach the heights that Rockstar wanted to reach. Hell, it didn’t even need Guarma, yet the studio’s insistence on meeting such seemingly arbitrary and ultimately pointless benchmarks for size and detail contributed to a crunch culture so infamously brutal we’re still talking about it today, nearly a decade later.
As with painting, what games lose in visual and technical polish they’ll make up for in conceptual depth. Before the camera, painting was all about what was being painted, not how it was painted. Painting today, on the other hand, typically isn’t about the painting itself so much as how we, the viewer, interact with it: how artists manipulate shape and color to draw attention to the ways culture and brain chemistry quietly shape our perception. Just as the Mona Lisa makes you forget you’re looking at a painting as opposed to a real person, many triple-A games today want you to forget you’re playing a game and make you feel like you’re inside some kind of Hollywood blockbuster instead. Tomorrow’s games – rejecting the immersive potential of AI – will want to make it clear that you are, in fact, playing a game: something invented, constructed, and meta.
While many creatives fear that AI will put them out of work, there are convincing reasons to believe this will not happen – at least, not to the apocalyptic levels they anticipate. Paul Downs, a CG animator and animation teacher in Florida who I spoke to for a different, upcoming IGN article about animation, told me he believes that “AI slop will cancel itself out.” The more low-quality AI-generated content gets released into the world, he reasons, the more people will hunger for genuinely human art. This would be true of any artform, games included.
Two other seasoned artists I spoke to – Dariush Derakhshani and Sam Nielson – both agreed. Derakhshani recalled a time when his supervisors used GenAI to storyboard a film, only to give the job to a human when the results proved completely unusable. Comments by EA’s Andrew Wilson or DreamWorks co-founder Jeffrey Katzenberg – who once predicted that AI would slash labor costs by up to 90% – are, as far as Derakhshani is concerned, nothing but “empty posturing to inflate stock prices.”
Nielson, who teaches at BYU and worked as a designer on the Game Boy Advance game LEGO Bionicle: Quest for the Toa, echoes the results of the various Game Developer surveys, arguing that “the inherent complexity of both design and storytelling” and the “specific psychological criteria” for how audiences respond to those things make it difficult for AI to fully automate creative processes. “A retired colleague who taught 3D animation used to say that students who were hoping to find the button that makes animation easy were making a mistake,” he told me, offering an anecdote that – to my ears – rings true for gaming also. “Because once that button is invented, everyone will push it.”
Then again, perhaps not everyone. In a future where many companies, big and small, will turn to AI to automate significant and in some cases just about all parts of their production processes, we can be sure that a number of creative and ambitious developers will keep on making games that reflect and relish in the unmistakably human craft behind them.
Tim Brinkhof is a freelance writer specializing in art and history. After studying journalism at NYU, he has gone on to write for Vox, Vulture, Slate, Polygon, GQ, Esquire and more.
Google might reshuffle search results to try to dodge fines in the EU