I've experimented with them, I think current models could replicate (with just a few prompts and on the free tiers) the "interesting mechanic rather than complex levels" kind of content I saw on Kongregate circa 2010.
Right now, they are pretty bad at balance, so you'd have to do that yourself. Last I tried they were awful-bordering-useless at level design, you must do that yourself.
I'd suggest not having any graphics come from an AI if you can avoid it. They're… annoyingly still in the uncanny valley.
By way of relevant example, here's one I made. Code and images were generated at least in part by an AI, mostly ChatGPT, the music was from a procedural generator I made myself in 2009 for shareware games I wrote back then. But I had to edit the images by hand because a significant fraction of the time "transparent background" an image with a white-grey checkerboard background, and all of the time "pixel art" was still a megapixel image. Oh, and the background didn't actually tile. And the achievements were even more generic. And the distribution of aliens in each level, and how weapons upgraded, just wasn't fun until I fiddled with it.
As a forewarning, there is a bug in the music player which can be painful to listen to when it triggers. I do not recommend vibe-coding something to parse and play MIDI, you should use something built into the browser like MP3 instead:
^Id like to add, if I might: Creating HTML wireframes to iterate back and forth with your agent of choice is a great way to work on UI/UX without the heavy token burn that comes with making iterations in the app code itself. Planning out backend and frontend in detail before implementation is the best way, I find, to save yourself from burning unnecessary credits.
As Ben points out, creating artwork and detailed assets to your liking is definitely most fun when you do it yourself. It's a great way to learn a new skill or two while you're at it.
I grew up playing flash games on sites like miniclips or addictinggames.
IMO there's no strong reason why games should or shouldn't be created using AI (entirely or even partially). There's a few software companies after this space but I think the issue you run into is monetization -- I'm probably never going to pay to play a flash game equivalent, and im pretty confident that's the majority of people which leaves ads and microtransactions as the dominating paths to profitability.
I did too, and I'm pretty sure you're right... it won't be a profitable venture itself... but if you make a few games and keep players then perhaps you can sell that side to others as a development 'studio'
In my eyes generative AI opens a lot of doors for people (such as yourself) that have a creative spark and wish to create/develop without monetisation being a concern or priority.
Creating something that can be beneficial, interesting, or simply just a way to pass by time, without financial motive, is in many if not most cases truly rewarding in and of itself. Especially when you consider that most people would prefer no ads and/or micro-transactions, which is often not the case in today's world.
One obstacle to be aware of, if you ever find yourself working on something you become passionate about and would like to release, is that Apple (Possibly other mobile app marketplace providers as well) are constantly introducing more ways to crack down on AI generated apps (In slang terms: AI slop).
Just some of my thoughts on a most relevant topic that you have brought up.
Definitely interesting to read about any apps you have recreated/attempted to, if you're willing to share:)
I've experimented with them, I think current models could replicate (with just a few prompts and on the free tiers) the "interesting mechanic rather than complex levels" kind of content I saw on Kongregate circa 2010.
Right now, they are pretty bad at balance, so you'd have to do that yourself. Last I tried they were awful-bordering-useless at level design, you must do that yourself.
I'd suggest not having any graphics come from an AI if you can avoid it. They're… annoyingly still in the uncanny valley.
By way of relevant example, here's one I made. Code and images were generated at least in part by an AI, mostly ChatGPT, the music was from a procedural generator I made myself in 2009 for shareware games I wrote back then. But I had to edit the images by hand because a significant fraction of the time "transparent background" an image with a white-grey checkerboard background, and all of the time "pixel art" was still a megapixel image. Oh, and the background didn't actually tile. And the achievements were even more generic. And the distribution of aliens in each level, and how weapons upgraded, just wasn't fun until I fiddled with it.
As a forewarning, there is a bug in the music player which can be painful to listen to when it triggers. I do not recommend vibe-coding something to parse and play MIDI, you should use something built into the browser like MP3 instead:
https://benwheatley.github.io/JS-game-engine/JS%20game%20eng...
More recently I've made a tower defence game, but I've not uploaded it anywhere.
^Id like to add, if I might: Creating HTML wireframes to iterate back and forth with your agent of choice is a great way to work on UI/UX without the heavy token burn that comes with making iterations in the app code itself. Planning out backend and frontend in detail before implementation is the best way, I find, to save yourself from burning unnecessary credits.
As Ben points out, creating artwork and detailed assets to your liking is definitely most fun when you do it yourself. It's a great way to learn a new skill or two while you're at it.
I grew up playing flash games on sites like miniclips or addictinggames.
IMO there's no strong reason why games should or shouldn't be created using AI (entirely or even partially). There's a few software companies after this space but I think the issue you run into is monetization -- I'm probably never going to pay to play a flash game equivalent, and im pretty confident that's the majority of people which leaves ads and microtransactions as the dominating paths to profitability.
I did too, and I'm pretty sure you're right... it won't be a profitable venture itself... but if you make a few games and keep players then perhaps you can sell that side to others as a development 'studio'
Or just be in it for the love of the game.. :)
People will play AI-develped games, the problem is marketing them to a playerbase that is very vocal and hostile about it.
It's 2026. Every game has AI generation somewhere on the process
No
Who do I work for, the fbi or the cia, please?
In my eyes generative AI opens a lot of doors for people (such as yourself) that have a creative spark and wish to create/develop without monetisation being a concern or priority.
Creating something that can be beneficial, interesting, or simply just a way to pass by time, without financial motive, is in many if not most cases truly rewarding in and of itself. Especially when you consider that most people would prefer no ads and/or micro-transactions, which is often not the case in today's world.
One obstacle to be aware of, if you ever find yourself working on something you become passionate about and would like to release, is that Apple (Possibly other mobile app marketplace providers as well) are constantly introducing more ways to crack down on AI generated apps (In slang terms: AI slop).
Just some of my thoughts on a most relevant topic that you have brought up. Definitely interesting to read about any apps you have recreated/attempted to, if you're willing to share:)
Code yes. Art no.
[flagged]