There is currently a concerted and not-so-subtle campaign underway to try to manufacture consent for the use of AI in game development.
At its core, this is about presenting AI as a fait accompli. Every week or so we get a new report making dubious claims about almost every development studio already using AI. The message to developers is “you’ll be left behind if you don’t use it too”; to players, it’s “this is already done and there’s no turning back”. Consume like good little consumers. Shut up and eat your slop.
This is, of course, a microcosm of broader efforts by tech giants to convince the world of the inevitability of AI permeating every facet of life and work. It follows the same playbook and rests on the same foundation of dubiously conflating many different tools and systems, including deep learning models that predate LLMs by many years, under the single banner of “AI usage”.
In game development, that can mean all manner of rhetorical sleight of hand. Anything from programmers using Copilot autocompletions, to 2D artists using a background removal tool in Photoshop, to a project manager using the meeting summary tool in Zoom, can all end up counted as a tick in a box and a “studio that’s using AI”.
A good rule of thumb is that when someone pops up every five minutes to dramatically insist something is a fait accompli, that’s a pretty sure sign that it’s precisely the opposite. Nobody feels the need to convince the world of a new reality when it’s actually real.
None of this is to say that plenty of developers and studios aren’t experimenting with AI tools or considering how they fit their workflows. Games are a cutting-edge, high-tech industry as well as a creative one. To the extent new AI tools are useful, reliable, and effective, they’ll be adopted. People will automate writing tons of boring templates or testing code, and welcome smarter tools for automating parts of the development pipeline for game assets. Nobody in a high-tech industry will turn down genuinely better versions of IDEs or image editors; that doesn’t mean they want to fire their art department and replace it with half-assed prompting.
“Nobody in high-tech will turn down better versions of IDEs or image editors; that doesn’t mean they want to fire and replace their art department with half-assed prompting”
Even as developers explore useful tools washed ashore amid the storm tide of hype, they’re conscious of a massive factor looming over all of this. There’s a reason why many dubious, anonymous surveys claim loads of development studios use AI, yet few actual studios proudly say so. Two reasons, really: first, as mentioned, many studios would probably contest that what they’re doing constitutes “using AI in development” in the way people actually think of it. The second is that they know that their customers despise the idea of AI in games – a stance that has only become more deeply ingrained among consumers as AI boosters’ claims of inevitability have grown more shrill.
Consumers really don’t like AI, and the challenge for those hoping to change minds is that the reasons for dislike are varied and largely valid. Some dislike the aesthetic of AI-produced work – many can smell it from a mile off, and it’s shocking how many ordinary folks have a viscerally negative reaction to seeing AI artwork or hearing AI-created music. Some hate the ethics of AI, with good reason; the argument for using training data to create a language model looks thin when it creates a model that mimics people’s artwork, voice, or image wholesale. Others rail against data centres, economic impacts, or just hate the unlikeable billionaire figureheads of the AI wave itself.
Any attempt to normalise extensive AI usage in development will crash hard against that wall. Most consumers want art made by humans and will vote with their wallets to get it. They’ll also demand the right to make informed choices, as seen with calls on Steam not only to ensure that AI-created games are clearly labelled, but that consumers have a straightforward way to filter them.
AI boosters like Epic’s Tim Sweeney have railed against the labelling of AI-created titles, which is something of an admission in itself; if they were genuinely convinced that AI was poised to be embraced by the world as a valid and welcome creative tool of the future, they wouldn’t be so adamant that developers must be allowed to use it under the cover of darkness.
The situation around games developed using AI extensively on the Steam platform bubbled up again a couple of weeks ago when the Steam Next Fest, an event that’s meant to showcase hidden gems on the service, turned out to be riddled with lazily churned-out titles filled with AI assets.
Players demanded a strict filter to remove AI titles, a reasonable ask assuming “AI titles” can be defined meaningfully. For the broader industry, this highlights a problem with AI that’s rapidly growing out of control, with far larger consequences than any debates over the creative or ethical validity of AI tools. There’s an avalanche of slop bearing down on us – and nobody knows how to cope with it yet.
“Tools now churn out game after game filled with IP-infringing assets, letting those willing to pay hefty fees release several titles a week”
To see a miniature example of the chaos coming, look at retro gaming spaces now. In recent months, a niche community devoted to “decomps” – decompiling old, usually abandoned games so they can be re-implemented on modern hardware – has been overtaken by a gigantic wave of AI-driven, automated decompilations of many games, mostly still commercial titles.
These vibe-coded recreations sometimes mash bits of different games in ways that probably have corporate IP lawyers tasting colours and seeing entire new planes of reality. They remain largely a curiosity, yielding more hours of TikTok videos than playtime. However, they have drowned out that corner of the retro gaming space, likely destroying it forever – and in the process, pointed a grim finger at what’s coming for other parts of the industry in the near future.
The bar to making games was already much lower than in the past, and in general, that lowering has been a good thing. High-quality development tools and the democratisation of distribution platforms have created whole new industry segments and opened doors for exciting, diverse new creative voices; it’s not even worth thinking about how creatively barren videogames would have looked over the past decade without that.
What’s happening now, however, is that the bar to making something that passes as a game has dropped below ground level. Tools now churn out game after game filled with IP-infringing assets, letting anyone willing to pay a hefty fee release several new titles a week.
This echoes one of the problems encountered by corporations that adopted generative AI in the past couple of years. It’s called the “slop grenade” – meaning that people using AI produce massive amounts of reports, emails, and other output materials (no doubt feeling terribly productive in the process), but all of this constitutes a hand grenade of AI slop that explodes on someone else’s workload, because eventually someone has to go through everything they’ve produced.
This is exactly what’s happening to gaming storefronts – already on Steam – a massive increase in quickly produced, low-quality games given a vague sheen of professionalism by models trained on countless professionally produced video games.
Slop itself isn’t a new problem, of course; anyone who watched the early promise of mobile gaming disappear under a torrent of slop knows that, as does anyone who’s ever dipped a toe past the first page of Nintendo’s online store. Reskins and asset flips are commonplace; many marketplaces teem with games that barely function past the first few screens or that shamelessly aim to rake in some coins from unsuspecting consumers misspelling the names of legitimate titles in their searches.
What’s new is the sheer scale of what’s coming – a scale at which even if every slop title only ekes out a few quid here and there, their cumulative effect could still crash discoverability, destroy marketplaces, and leech massive amounts of revenue away from real, genuine games.
Some creators are sounding the alarm about this; Smash Bros. creator Masahiro Sakurai mentioned in a recent interview that he already sees AI accelerating a flood of unoriginal titles that copy other people’s ideas and fill up storefronts with games that can never be profitable. It was interesting to hear this from a Japanese developer in particular, given how positive coverage of AI has generally been in Japan compared to in Western countries.
Solutions can’t come from creators; only people who can actually close the gates against the flood of slop, and it’s unclear whether they have the will or means to do so.
The system right now will not cope. Platforms like Steam have largely given up on earlier attempts to gatekeep access – people hated when Steam only allowed invited titles, and still hated the greenlight system. So we’ve ended with what is essentially a free-for-all where any game can launch on Steam if it pays a small fee. That worked, to a degree, for a while. It’s about to stop working.
“We need to stop talking about AI as an inevitable future and start talking about how to clear up the inevitable mess it’s going to make”
It will stop working because even when Valve stopped holding the gate shut, a gatekeeper remained: making a game, even a bad one, took time, effort, and knowledge. If that gatekeeper is also gone, then we’re about to see what a truly open market looks like for the first time in the medium’s history. I’m not sure that’s what anyone actually wants, least of all real creators who will fight to keep their head above this rising tide of slop.
In this, at least, creators and consumers will find themselves allied. The market is making it pretty clear that it doesn’t think doing nothing and embracing the slop is an acceptable answer, and platform holders like Steam (and console platform holders, and mobile storefronts) will have to respond to consumer demand eventually, even if they won’t respond to creators’ pleas.
How to fix the problem is a bigger, thornier issue, and we’ll no doubt see some silly fighting-AI-with-AI misfires along the way. The starting point will be a shift in the tone of discussion; we need to stop talking about AI as an inevitable future and start talking about how to clear up the inevitable mess it’s going to make.