Game Oracle, a gaming industry research platform, has published a two-part data study on how developers use generative AI in Steam games and how that choice affects early sales. The study examined all 16,554 titles released on Steam between January and November 2025 and sifted through their mandatory AI disclosure statements.
The study found that about 21% of those games indicate some form of AI use, with the monthly share peaking at around 25% in September 2025. (Again, this study ended when 2025 did, so these figures are expected to be much higher now: around 40%.) Most teams use AI for art, graphics, and coding, but it’s less commonly used for translation, audio, or storytelling. AI also appears across genres and art styles, not just in text-heavy or niche projects.
The Game Oracle team also examined what AI use means for commercial performance, analyzing 9,879 paid games released between January and October 2025. Using first-month Steam review counts as a proxy for sales, the study found that games using AI receive about 53% fewer reviews in their first month than similar titles that don’t use AI. Games created with AI have lower median review scores, more zero-star launches, and slightly lower positive ratings among titles with at least 100 reviews.
Credit: Game Oracle
The authors argue that AI might have a neutral effect on low-budget beginners who already face poor sales, but might hurt potential hits from experienced teams by feeding an “AI stigma” and reducing player trust. They say AI can still be a useful tool, but studios should wield it carefully rather than use it primarily to cut costs.
“…due to the unmeasured effects of marketing, skill, resources, and experience, our results could still be biased,” Dr. Ross Burton, head of product and data at Game Oracle, wrote. “But we are certain of one thing, for many developers the effect of using AI is currently a net negative.”
