The numbers, without the hype
Roughly half of game studios now use AI somewhere in production, and more than 7,300 games on Steam disclose some AI usage, double the 2024 figure. At the same time, industry surveys put developer sentiment sharply the other way: around 52% of professionals say generative AI is harming the industry. Both things are true at once, and the gap between them is the most useful thing in the data.
Where AI is genuinely used
Adoption is concentrated in research, coding assistance, prototyping and administrative work. It is lowest in the creative and technical-art tasks that decide whether a game actually feels good. That distribution matters: AI is compressing the unglamorous middle of production, not the parts a player notices. Around 36% of professionals report using generative tools personally, with ChatGPT the dominant one.
What it does not do
It does not design a satisfying core loop. It does not work out why a mechanic is boring on the third playthrough. It does not hold a project's architecture in its head across six months of scope changes. The studios shipping well with AI use it to remove drudgery so senior people spend more time on judgment, not so they can employ fewer senior people.
What this means if you are hiring a studio
Two questions are worth asking any partner. First: where do you use AI, and where do you refuse to? A studio with no answer either has not thought about it or is not being straight with you. Second: who owns the output? If AI-generated assets are in your game you need clarity on provenance and licensing, because the legal position on training data is still unsettled in several jurisdictions.
The honest risk
Cheap AI-heavy production is real and some of it ships. It also tends to produce games that look competent and feel generic, because the parts AI is worst at are exactly the parts that create distinctiveness. If a quote looks impossibly cheap, ask what is being generated rather than made.
How we use it
We use AI for what it is good at: boilerplate, tooling, first-pass code and research. Not for the design decisions, art direction or optimisation work that determine whether your game is worth playing. Every asset we deliver is one we can account for, and you get 100% source-code and asset ownership either way. See AI in game development.
Related: AI game development · Choosing a studio · Unity development
Quick answers
Will AI replace game developers?
Not on current evidence. Around half of studios use AI, but mostly for research, prototyping and coding assistance rather than final creative or technical-art work. It is compressing production overhead, not replacing the judgment that makes a game good.
Do you use AI to build client games?
We use it for boilerplate, tooling, first-pass code and research. We do not use it for design decisions, art direction or optimisation. Every asset we deliver is one we can account for, and you own all of it.
Is AI-generated art safe to ship in a commercial game?
It depends on provenance and the jurisdiction you publish in, and the legal position on training data is still unsettled. Ask any studio to be explicit about which assets are generated and what licensing applies before you sign.
Does AI make games cheaper to build?
It reduces some production overhead, which can lower cost on the mechanical parts of a build. It does not reduce the design, art-direction and optimisation work that decides quality, so a dramatically cheaper quote usually means less of that work rather than more efficiency.
Related guide
What GTA 6 means for game developers →Building something, with or without AI?
Send us the brief and we will tell you honestly where automation helps your project and where it would cost you quality.