The way that valves AI tag works is kind of a problem.
There is no subtlety to it at all, if you use AI in any capacity during the development of the game you need to declare it via that tag yet all the tag then does is say “AI in this game”, but there’s a big difference between having the AI develop the entire story or produce all of the artwork, and having AI write boilerplate camera controls for a farming simulator.
I agree that having more degrees of usage would be useful, but erring on the side of caution and declaring any AI use as a first step is better than doing nothing.
Okay so there is this whole arguement going on about The Altars how apparently a tiny piece of background art has AI generated text in it. Personally I feel that’s absolutely fine, as otherwise it would have just been Lorem Ipsum, and really doesn’t need to be declared but technically, under the strictest interpretation of that tag, it should be declared even though you can’t even see it unless you zoom in.
I would very much like valved actually come up with a concrete policy rather than a vague one-line statement.
E.g. Does using Intellisense need to be declared?
Would syntax highlighting?
Does using copilot to code count as “made with AI” too?
Of course, that’s why we need better guidelines. It’s like beauty ads that have to declare they used Photoshop. Every photo is edited if you don’t make it clear what you mean
Why should something not be disclosed just because its common?
It builds indifference to the disclaimer when it’s too general. The California cancer label is a good example.
But its not too general. Steam allows you to give a description of the use of AI
The problem is you end up with the tag nonetheless.
The description doesn’t apply to the label, no matter how much explanation you provide you’re still going to devalue your game with the AI label so why would any developer admit to that?
The whole thing is just mind numbingly stupid.
Whoever thought this up needs to get out more and actually experience the human condition.
I didn’t say that. It should be more specific to have any meaning to the consumer.
But it has meaning to some consumers. Not everyone can tell that an image has been majorly edited or created using a program created to replicate pictures
Because it becomes meaningless noise instead of useful information.
Even if it is ignored by a lot of people, its better than not knowing at all
I’m not saying there shouldn’t be a disclosure, but an uncertain threshold that might be as low as “a developer accepted a copilot completion suggestion one time” isn’t useful. You just end up with a prop65 situation where it’s slapped on everything and basically meaningless.
Yeah, I suspect the AI tag should apply to even more games then.
Hence the problem
I figure if people can’t be bothered to develop the games then I can’t be bothered to play them either.
I always wondered if I could play Skyrim but with an option to say or ask custom things to NPCs
I think someone made a mod for that because of course they did.
They absolutely did, awhile ago too. It’s the Mantella mod. Used to be a nightmare to set up but now it’s good. Pretty fun talking to characters.
Is it bad that I know what the Mantella is in Elder Scrolls lore?
There was a question about Daedric Princes in a trivia game I went to recently. They gave four of them and you had to match what they were the prince of. I felt confident about two but only got one right. Of course I got Shegorath right.
What was the one you messed up? I’m gonna guess either mephala or meridia since they are kinda weird, either that or the answer was Magnus who is technically a daedra though not usually lumped in with the rest.
I thought Sanguine or whatever was the murder one but I remembered wrong. They gave the four categories and we had to match them. I thought Malacath might be based on Malakith from War Hammer and that seemed pretty hedonistic but it was a 50/50. (Well, I thought it was a 50/50 lol.) The last was Mephala who I thought might be lies. And if you know the lore well you probably already knew Mephala was the fourth they mentioned lol.
I think the biggest problem is that steam is like 80+% shovelware and it’s no surprise that a lot of those are using a bunch of AI generated “artwork.” IMO it’s no worse than a shitty asset flip and as others have pointed out, there are a lot of really cool things you could do with generative AI in game dev that aren’t just slapping shitty pictures all over your product, and this doesn’t capture the nuance. I would also assume that this number is lower than reality since it relies on tagging, and nobody is accurately tagging shitty scam games with less than a hundred downloads.
what I want with AI games: Free conversations with NPCs who react to your actions.
what I don’t want, endless slop
Same, so used as a tool to help textures load better, to make the game function better, great.
If you pay an actor for their voice, why not ai infinite dialogue? Nobody is losing work because a human literally cannot do that job.
without that tech they would have hired a voice actor anyways, or what would be better, get a voice actor, create a character, have him voice a shit ton of lines with different emotions, timber, whispers… fine tune a model in that character, that would make every character sound unique and much less robotic.
How is it possible to hire a voice actor to make up infinite lines based on my specific questions? For every player? It isn’t
not infinite lines, just enough to find tune a model for a specific character.
also, i mean proper voice actors, the ones that can make countless distinct characters with different voices and accents.
i don’t mean buying someone’s voice, i mean hiring a voice actor to make up a specific character and have ownership of that character.
what is the appeal of talking to an NPC that uses chatgpt to respond? you would get the same experience talking to a cat or a houseplant
instead of writing pages of dialogue, write a lot of back story, personality, interests, knowledge, info they have, quests they have to share, sample of how they talk…
fine tune models… this way each character would sound unique, rather than standard chat gpt.
a good prototype would be about a village with about a dozen of NPCs.
no thanks
another use, draw assets for a age of empires like game. then generate a diffusion model on them. now you can make rows of houses and non of them will be identical and all will fit in the art style.
same things with textures, no more repeating textures.
Yes pls
I don’t know if you have heard about the game AI2U: With You 'Til The End that’s basically a Escape the room game where a Girl kidnapped you and you need to escape. You can talk to the AI girl about any bullshit, interact and show them items in your inventory etc. If you make them upset they will maybe kill you, or they can like you so much that they will help you escape. I had a lot of fun with this game.
If I see some AI Slop thumbnail for your shit ass ripoff game then you can bet your ass I’ll never play it let alone ever pay money for it.
The conversation around gen AI seems to go to putting people out of work or replacing tons of human effort, and I’m sure some companies are led by people with those naive dreams, but that My Summer Car example is exactly where my head goes when I think what the future of the technology is. It’s artwork that ought to be there, because the scene demands that there’s art on the walls, but what that artwork is basically doesn’t matter, so if gen AI can get the job done cheaply, it’s probably the right tool for the job. However, I’d have thought that the scientist portraits in Jurassic World Evolution were another prime use case for it too, but people rioted over that one. Even if it’s a good tool for the job, if it’s poison in the marketplace, it’s no longer a good tool to use.
I find this outlook to be pretty sad. The idea of chunks of your art “not mattering” and just being there as filler.
One of the joys of creating artwork is that during the process of creation you are actively figuring out what is important. Perhaps you start out creating a simple texture just to have something on the walls, and in the process you realize there’s an equally simple yet creative way for you to tell a little story with that wall. Something most players will never notice but a year from release gets thrown in “small details you missed” compilations.
It may be that the idea you came up with for that wall goes on to influence the main story, and spur on a totally different and more interesting game than you initially imagined.
A lot of non-artists have this concept of art, where it forms completely in your head in a single burst, and then you just have endure the tedious labor of constructing it. I think that’s why people are so easily persuaded by the ‘promise’ of AI. They think it’s just making the boring parts easy. But in reality it’s making the creative parts boring
I think when you’ve got a small enough team making something as multifaceted as a video game, there will be parts of it you find boring and relatively unimportant. If you can make it cheaper, you get that much closer to the possibility of breaking even. Parts of this can scale up to larger projects, but in the end, this is a matter of choosing your battles. There’s an adage that’s something like, “Your game is never done; you just stop working on it,” and the sooner you can stop working on it while still delivering a product that people are interested in, the more sustainable the whole endeavor becomes. Chunks of it will be filler or less important than other chunks, always. It’s why there’s a Unity and Unreal asset store; and why you can hear the same sound library used in Devil May Cry, Soul Calibur, and Dark Souls menus. Those parts of the game were less important to be specifically crafted for these games, and they chose other battles to care more about.
Eh, the small team argument doesn’t really carry any water I think. Some of the most beloved indie games of all time have simple, geometric graphics. Thomas Was Alone even managed to tell a tear jerking story between characters who were monotone squares and rectangles.
Using AI to totally gloss over some of your most basic creative questions, such as “what are my capabilities?” And “What can I do given those limitations?” Isn’t going to lead you to a better product. If something is truly that unimportant it can be arranged trivially or cut. Even choosing to cut something is an inherently creative decision; another layer of the process which is lost if you train yourself to reach for AI to implement something that suits your first whim.
The asset store angle is also not really comparable. You’re still collaborating with another artist. We could ride this train all the way down to you didn’t personally mine the silicone for the computer you personally designed if we felt like it. It’s disingenuous and ignores the material differences between these technologies.
In summary, I basically think that you are narratively framing this as something that empowers the little guys, but I disagree that it is actually doing so in practice. It’s a product that’s only on our minds because of a massive concerted effort on the behalf of mega corporations whose explicit goals are to rob and disenfranchise us
It does empower little guys. It empowers everybody, for good or ill. That’s a what a tool is. I can hammer a nail, I can hammer a nail that shouldn’t be there, or I can hammer a person in the face. Is any of that the hammer’s fault?
Gen AI is not a hammer, it is an auto-construction machine that removes you from the process of building. It doesn’t empower people, it sidesteps them.
Oooohhh Grooosssssss! It’s gonna fucking overtake the real content so fast, now. jfc, how do we even sort them out if the “creators” don’t follow the disclosure rules?
Read reviews. Examine promotional materials. Discuss with friends.
So at the end of the day we all have to go digging through muck. Horrible.
It doesn’t take very long to look at ratings. If it’s crap it’ll have crap ratings.
A lot of great games get middling reviews, but I’m expected to parse whether or not something contains slop by a glance?
I mean that’s exactly how it’s always worked. What’s the difference just because the AI exists
No, theres a fucking difference, mate. AI has empowered the worst people to make shit they otherwise couldn’t have, and it has degraded the works of anyone with some level of skill.
Yeah because there hasn’t been a bunch of asset flip slop before. All it’s changed is that they have new tools to make it but there was always crappy stuff on Steam
You need to train your ability to spot AI.
AI art has a very distinctive style. Weird shadows, impossible architecture, and having a blatantly incorrect number of fingers are dead giveaways.
AI text tends to talk at you rather than with you. It has difficulty remembering context, so it tends to forget what you said 10 lines ago.
Steam should combat shovelware whether it’s AI slop or human slop
Procedural generation of content in games is by no means a new thing. Even if the end state isn’t completely procedurally generated, odds are a version of the asset was initially and a human touched it up as necessary. When you’re talking about large asset sets (open world and/or large maps, tons of textures, lots of weapons, etc) odds are they weren’t all 100% hand made. Could you imagine making the topology map and placing things like trees in something like RDR2?
That’s not to say all this automation is necessary a good thing. It almost feels like we’re slowly chugging through a second industrial revolution, but this time for white collar workers. I know that I tell myself that I would rather spend my time solving problems vs doing “menial” work and have written a ton of automation to remove menial work from my job. I do wonder if problem solving will become at least somewhat menial in the future.
Is procedural generation part of what they must disclose as being AI generated, tho? The assets used usually are still made by hand, even if all the tree and rock (and whatever else) assets were placed procedurally using RNG or some sort of algorithm.
Procedural generation is just an earlier form of AI that’s been demystified and commercialized as a positive thing. The current generative tools aren’t at that level of adoption yet.
Procedural generation is just an earlier form of AI that’s been demystified and commercialized as a positive thing.
That is such a broad generalisation that it makes the term useless.
Is Minesweeper AI then? It uses procedural generation to generate the play area. Is Minecraft chunk generation AI?
Is perlin noise AI?
I wonder if games with UGC report they have AI content. (Games that allow for outside assets and code)
Are algorithms the same as AI?
I have an acquaintance who is a lead Dev at an Indie studio where he is developing and training an NPC behaviour engine with thousands of responses and actions. Think fallout or mass effect response wheel, where 2-4 dialogue choices have 2-4 outcomes, but instead you can tell the NPC anything and it will have a different response. Or it will do different things whether you hand it a book, give it book, throw a potion at it or cast a healing spell on it or hug it. It could also change tactics if you tried to snipe it vs if you went at it melee. All of these are trained and accounted for and made in a way where it can be built into any game using a certain engine. And this is just aimed at generic npcs, not companions.
So if this is what disclosure of the use of generative AI means, I’m not against it. I think there is nuance to what can be done with it. Using final art assets? It’s theft. Writing? Theft. NPC behaviour? Definitely not.
Strange to not qualify the last one as theft. If it’s out putting code, it’s from the same kind of training set. If it’s out putting character responses, they’re from that same literary training data.
Open-source training texts intended for pairing with your intended style of output have been around for far longer than OpenAI has been grifting data from the entire Internet and collected book works. It came across like that’s what they’re using, not some shit off HuffingFarce that was built off of AO3 and Harry Potter.
How many of the ~6,818 titles now disclosing generative AI use were already on Steam in 2024?
I.E. are a lot of these just games that had already been released, updating their disclosure statements based on Valve’s new rules?
The article says 1/5 games released this year use it. I’m not sure if ~34,000 games have released on Steam in the last year
It is a little insane how many games release on any given day. On July 15, 2025, 150 “titles” (of which 78 are actual games, not demos or DLC) were added to the Steam store. I would guess that their data includes all titles, but even just 78 real games on what should be a slower-than-average random Tuesday could totally contribute to 34,000 games released in a year.
Honestly, maybe I’m an old fart, but I refuse to knowingly buy games if they use AI instead of paying talented people to create works of art.
What if talented artists use AI to enhance their original work?
An interesting use case for me in programming has been prototyping. Stuff I otherwise wouldn’t have the time to experiment with suddenly becomes something feasible. And then, based on what I learnt while having the AI build the prototype, I can build the actual thing I want to build. So far, it has worked out pretty nicely for me.
I read a story recently about how a graphic designer realized they couldn’t compete anymore unless they used generative AI, because everybody else was. What they described wasn’t generating an image and then using that directly. They said that they used it during the time when they’re mocking up their idea.
They used to go out and take photographs to use as a basis for their sketches, especially for backgrounds. So it would be a real thing that they either found or set up, then take pictures. Then, the pictures would be used as a template for the art.
But with generative AI, all of that preliminary work can be done in seconds by feeding it a prompt.
When you think about it in these terms, it’s unlikely that many non-indie games going forward will be made without the use of any generative AI.
Similarly, it’s likely that it will be used extensively for quality checking text.
When you add in the crazy pressure that game developers are under, it’s likely that they’ll use generative AI much more extensively, even if their company forbids it. But the companies just want to make money. They’ll use it as much as they think they can get away with, because it’s cheaper.
What I dread is a game lengthening dialog using AI. Some folks mistake quantity for quality, and make their games unbeatingly tedious. Just like games that lean heavily on procedurally generated content.
Yep, not excited for Starfield generated planets type of deal when it comes to dialogues and such.
My personal issue with the idea of “infinite NPC dialogue” is that it defeats the purpose of minor NPCs. They’re just there to give you a nudge in the right direction or give flavor text (“Bandit activity sure has been picking up!” or “The king? He’s probably in his castle to the west.”). Turning them into a chatbot just means a player potentially spending all their time there with nothing to gain that they couldn’t get from Character.AI instead of playing the game.
I’m also curious about the implementation. AI API use isn’t free so you’d likely be requiring players to pay if they don’t meet the hardware requirements to host locally.
Funnily enough, I’m excited for new dialog in video games using generative AI. It would be nice for random NPCs to not have the same 3 recorded voicelines, but to actually change what they say based on what’s happening around them.
But that’s obviously a limited use for AI. It should definitely not be used to lengthen the game and clutter up storylines as you’re kinda describing.
