The slop flooding Steam isn’t a tool problem. It’s a taste problem. AI didn’t lower the floor for making games, it removed the excuse, and those are not the same thing.
Here’s the number that made me want to write this. In this year’s GDC State of the Game Industry survey, 52% of game professionals said generative AI is having a negative impact on the field. A year earlier that figure was 30%. The year before that, 18%. So in two years the industry went from mildly skeptical to outright hostile, and the people leading the opposition aren’t the folks you’d dismiss as behind the times. Visual and technical artists came in at 64% negative, game design and narrative at 63%. The people closest to the craft are the angriest.
I want to take that anger seriously, because it’s earned. Those artists are watching studios cut headcount and backfill with a text box. The quality collapse they’re describing is real: by March, more than 7,300 games on Steam carried an AI-disclosure label, and a good chunk of that wave is the thing the discourse now calls “gameslop,” a store page of generated filler with no hand on the wheel. If your livelihood is drawing, and the market’s answer is “we’ll generate it for a tenth of the cost and ship it half-finished,” 64% negative is a rational read of your situation, not a failure of imagination.
So I’m not here to tell the artists they’re wrong to be scared. I’m here to argue that “AI equals slop” quietly swaps two different claims for each other. One is about jobs and markets, and it’s true. The other is about the tool itself, and it isn’t. The slop isn’t coming from the tool. It’s coming from whoever’s holding it.
The temptation is real, and I feel it too#
I’m not a studio. I mess with Godot side projects on weekends and run tables in Foundry VTT. Small stakes, nobody’s job on the line, just me and a scene tree at 10pm. And from that seat, the temptation the survey is describing is completely legible, because I feel the exact same pull.
Say I’m blocking out a little top-down game. I need tiles, props, a UI kit, some sound. The honest 2026 move is to open a generator and ask for a thousand assets, and it will hand them over before my coffee’s cold. Placeholder art becomes final art because it’s there and replacing it is work. A dungeon fills with forty variations of a crate I never once looked at closely. Each one is fine. The pile is slop. Not because a model made them, but because I shipped a pile I never curated, and “it’s already generated” did all my deciding for me.
That’s the whole trap in one move. The cost of making an asset dropped to roughly zero, so the cost of not cutting one dropped too, and cutting is exactly the part that used to make the thing good.
This is the “AI amplifies you” thing, again#
I’ve written before that AI amplifies whatever you already are. It’s leverage, and leverage is indifferent: it multiplies clear thinking and sloppy thinking at the same rate. Game dev is just the most visible place that’s ever been true.
The person making something good with these tools and the person making gameslop are often using the same tools. Same generators, same models, same afternoon. The difference isn’t access and it isn’t skill with the prompt box. It’s taste, restraint, and knowing what to cut. One of them generates a thousand assets and ships forty. The other generates a thousand and ships all thousand, then wonders why the store page feels like a landfill. The tool amplified a decision. It didn’t make one.
Cheap generation didn’t lower the bar for a good game. It raised the value of the only thing it can’t do, which is judge its own output. When anyone can produce infinite raw material in an afternoon, the scarce resource is the person willing to throw most of it away. That person was always the difference. AI just stripped away every other excuse and left taste standing alone in the room.
The part my argument doesn’t fix#
Here’s where I have to be honest about the limit, because the counter is strong and dodging it would be its own kind of slop.
Even if every developer suddenly grew impeccable taste tomorrow, the flood is still real. Those 7,300+ labeled games sit in the same store, the same search results, the same “discover” shelf as everything made with care. Discovery is genuinely worse now, and a taste argument does nothing about it. Good taste governs what you ship. It has zero authority over the volume everyone else is dumping into the channel you have to be found in. That’s a market problem, a curation problem, a platform problem, and it lands on careful developers who did everything right and still can’t get seen through the noise. I won’t pretend the taste framing solves that. It doesn’t. It’s the argument for how you make something worth finding, not a fix for the finding.
But notice those are two separate fights. “The store is drowning in low-effort output” is true and awful and structural. “Therefore the tool makes slop” is the non-sequitur riding along with it. Conflate them and you end up swearing off an instrument because other people play it badly in a crowded room.
What I actually do#
The rule I hold, in a hobby where nobody’s making me, is small and boring. Every asset earns its place or it’s cut, and “the model already made it” is not a reason to keep anything.
The clearest test I know is a game jam. A jam game is one hand-tuned mechanic, tuned by a human who played it fifty times and felt where it dragged, wrapped in whatever art the weekend allowed. It’s often ugly. It’s frequently better than a store page of immaculate generated filler, because a person made a hundred decisions about that one mechanic and stood behind every one. Slop is the absence of those decisions dressed up in competent-looking output. The fix was never a better tool. It’s being someone worth amplifying, and then having the nerve to cut the other nine hundred and sixty crates.
