Waiting for the “Whoops, we ‘forgot’ to remove it”.
Sigh, TPTB are just intent on ruining The Expanse for us aren’t they?
Sure, once crunch times starts, you’ll have plenty of free time/people to audit the code and resources to determine if they were AI generated or not.
Fuck (this particular use of) AI.
Crude MS-Paint drawings are the best placeholder art. Scribbles can be made as fast, if not faster than AI. It shows the information it needs to show, but most importantly its incredibly obviously unfinished art that needs to be redone before release.
AI art looks pretty good at a distance without close inspection. You have to look closely and spend time to tell AI from art. Late in production when rushing for the deadline is not a point when you have the time to look closely at the assets, so AI placeholders will get missed.
Placeholders is a bad use case for AI.
And what you describe is probably the best case scenario, that the well meaning artist misses it in a rush.
In the real world much of the time I bet the artist is well aware and frustrated because the game already shipped (with no input from them or their boss) to make it by the end of the quarter and now the press assets all have some bland slop BS front and center instead of the real version staring back from their own monitor.
Welp. Waiting for a 75% off sale instead of day one purchase 🤷🏻
That’s fine, I’m not gonna play it.
if you generate graphics, story and stuff with llm and make them more pretty through human labour, you still have llm generated crud. Generated story acted by humans is still same as fully generated one. using generated slop as placeholders might be fine, but it most likely still influences what they actually make so still no.
if you are just starting something and have no idea about anything and no mental image about what is what at all, THEN using generated stuff might be okay as an example so you can get the idea what is going on, as long as you stop using anything generated as soon as you can.
llm and any datacenters involved with it need to go up in flames.
If we conveniently forget about how unethical all AI is, Slay the Spire shows that no game needs generative AI for “placeholder” content and Expedition 33 shows that it can and will slip through the cracks. Don’t. Fucking. Use. It.
Yep. I saw one slop picture from a new game that I was actually excited to buy, and even though it was “oh, we only used it as a placeholder and just forgot to remove after tehehe”, I immediately lost all the interest, and will not buy it now. Later I learned that the quests there are nonsensical and a bit disjointed, and the story is stupid, and I can tell you, I’m not surprised at all.
LLM is like that black goo, everything it touches gets corrupted forever.
Huh, weird, you weren’t supposed to see that in the final version. But we fired the guy responsible and we’re sorry that we got caught
If you’re using AI in your process, even early in development, it means the game is no longer 100% human made, so stop the bullshit.
Literally every game that’s made today is using AI as part of the development process.
Damn near every Dev has tab completion on in their IDE. Which is AI based.
==========
Edit:
I used a term, but I guess y’all were unable to infer meaning from usage?
Auto tab or w/e it’s called (Some products literally call it tab completion). Visual Studio was doing it around 2018 IIRC, it’s ML based, always has been. Modern versions of it are almost entirely LLM based
Tab completion is a table lookup and has been common for like 20 years. There’s no LLM needed.
I used the wrong term, but I guess y’all were unable to infer meaning from usage?
Auto tab or w/e it’s called (Some products literally call it tab completion). Visual Studio was doing it around 2018 IIRC, it was ML based, always has been. Modern versions of it are almost entirely LLM based.
No, it’s not that we were “unable to infer meaning” (gaslight much?), you’re just wrong.
Firstly, tab completion has been around for effectively ever, and way predates whatever VS Code may have been doing. Ctags indexing, for example, has been around since 1992.
Secondly, even if you want to move the goal post by talking about some specific implementation of ML based indexing, ML is not LLM.
Like I said plenty of products call this tab completion, and it’s context aware completion, or predictive completion. I used an overloaded term but I would have thought after my explanation you would have understood what I meant by this point. You’re continued explanation of classic tab completion is shows otherwise.
and way predates whatever VS Code may have been doing
Also I said Visual Studio, not VS Code. 🤦
Secondly, even if you want to move the goal post by talking about some specific implementation of ML based indexing, ML is not LLM.
I very specifically said that it was ML based, The word was indicates past tense. 🤦
“Modern versions of it are almost entirely LLM based.”
I don’t know how you managed to completely skip reading that last line?
Here we are though arguing over reading comprehension issues. Which honestly is pretty classic for the internet.
What do you mean “tab completion is AI based”? We have had tab completion for years before LLMs were a thing.
Apparently meaning from usage cannot be inferred here? Or you’re just being intentionally obtuse?
- predictive completion
- context based completion
A not insignificant number of products literally just call it tab completion these days, because tab completion in many products & IDEa is by default predictive completion, which is ML based. And these days, LLM based.
Fuuuuck. I don’t really care about this one, but now I’m worried they’re slopping up Dark Heresy too.
And just like that I added it to my ignored list on Steam
Same. Steam is so inundated with AI slop that I’m now following like a dozen different curators that flag AI usage, for the cases when the developers “forget” to fill out their AI disclosure field D: (which I’ve restyled to be red and on the top of the page)
I only know of AI Check which had to make a second account cause they already hit the limit with how many games they were marking as containing AI.
I went through my curator list just now; these are the largest/most active ones:
- Game uses Ai - 961 reviews - last update: 2026-03-27
- Is it AI? - 774 reviews - last update: 2026-03-26
- AI_games_flag - 1,998 reviews - last update: 2026-03-17
- AI Check - 2,000 reviews - last update: 2025-12-23
- The Curator Page I Made to Flag AI Slop - 383 reviews - last update: 2026-02-13
- Ai? No buy! - 343 reviews - last update: 2026-03-23
- Does This Game Use GenAI? - 193 reviews - last update: 2026-03-27
- NO AI #HumanArtists - 119 reviews - last update: 2026-03-25
- AI Check 2 - 86 reviews - last update: 2026-03-23
Thanks! Followed all of them, although not a fan of “Does this game use GenAI” adding games that they don’t think have any. I get it’s being informational, but at that point you might as well do like Half-Life just to say it doesn’t have any.
(X) Doubt
Some people just dont want my money. [Steps out into the airlock]
If your placeholder doesn’t stick out like a sore thumb, it’s a bad placeholder. There is literally no workflow in which temporary assets shat by AI would be useful.
They just want to normalize AI use until people don’t care anymore. And with the waste of resources this shit represents, I just hope this never happens.
Here’s a good example of placeholder art being very visible from Slay the Spire 2
workflow in which temporary assets shat by AI would be useful.
Collecting VC funding, particularly demonstrations or even gameplay captures.
You don’t see the use in an artist viewing an approximation of the finished product in order to see what can be improved?
Do you suppose all conductors just write symphonies in their heads and never have to hear them out loud before deciding they’re done? Would it be useful to replace the tuba with a placeholder of a duck quacking, or do you think they might want it to sound like a tuba even though it’s not the final product?
You don’t see the use in an artist viewing an approximation of the finished product
I don’t because that’s not what artists do.
Artists are not people who bring nearly finished projects over the finish line. And if your finished project does not look anything like your nearly-finished AI assets, what are you actually using them for?
You’re saying artists never look at their work and decide to change something. Ok, buddy.
Okay, well if you’re going to be like that, I’ll talk to someone else:
An artist’s job is to pull together research, resources, history, knowledge, opinions, their own fluency in the language of the medium they’re using, and a bit of inspiration, and turn that into something interesting, or cool, or flashy, or thought provoking.
AI generation, even for the concept phase, skips 90% of that effort.
You can’t fabricate something with AI and then re-make it by hand later because these are two halves of the same process. By the time the hands are involved, there is very little left for them to do.
That’s a narrow definition of an art that doesn’t apply to real life. Like it or not, Marcel Duchamp won the argument over whether or not he was an artist.
Marcel Duchamp is not working on video games.
Marcel Duchamp does fit my definition.I’m certainly more interested in his work than yours.
Even a cheap toy synthetizer can make something close enough to a tuba sound to get an idea of what it sounds like. Need something better? people make sound fonts for that.
But maybe it’s better to use generative AI to potentially have something close to the real thing, just so you can have huge datacenters consuming absurd amounts of power and water too.
What are ya, a shill for the toy synthesizer companies?!?! They just want to take jobs away from us hard working tuba players!!n
