Steam data shows AI disclosure cuts review counts by 53% and drags down ratings

Market data analyst Ross Burton, writing for Game Oracle, crunched numbers on nearly 10,000 Steam releases and found that disclosing AI use carries a measurable commercial penalty — fewer reviews, lower scores, and a harder road for developers who would otherwise have done well.

The sample covered 9,879 games released between January and October 2025, with spam, purely commercial releases, and free-to-play titles filtered out. Of those, 17.9% disclosed AI use. At a surface level, AI-tagged games had slightly fewer reviews and a median rating roughly 4% lower than non-AI games when looking at titles with at least 100 reviews. But the more striking number comes after controlling for publisher size, developer experience, and game type: AI-disclosing developers saw approximately 53% fewer reviews compared to equivalent non-AI developers. Burton notes that to explain away that gap as statistical noise, some unmeasured factor would need to nearly triple the odds of AI adoption while simultaneously causing a 22% drop in review counts — an unlikely coincidence. The effect was also more pronounced for higher-quality, higher-potential games; for low-quality releases, AI disclosure made no measurable difference at all.

The findings land in a complicated moment for the industry. Games like The Finals used AI extensively and still succeeded, which Burton frames as evidence that how AI is used matters as much as whether it's used. Clair Obscur won game of the year awards despite AI-generated placeholder assets, and Crimson Desert has already sold millions this year after doing the same. On the other end, the new Crazy Taxi's AI use has dominated its press coverage in a way that arguably overshadowed the game itself. The picture is further muddied by widespread undisclosed AI use — Epic CEO Tim Sweeney has pushed back on the idea of mandatory disclaimers, and Sony just days ago promoted AI tools as a way to "unleash the creativity of studios."

Burton's conclusion isn't that developers should avoid AI, but that they should use it carefully — the report's hammer analogy puts it plainly: useful tool, wrong approach if you swing it at everything. Given the industry's continued investment in the tech and the lack of any consensus on disclosure standards, the stigma penalty is unlikely to prompt a course correction at scale anytime soon.