One in Three New Steam Games Now Carries an AI Label, But It Rarely Pays Off

The share of Steam games labeled as made with artificial intelligence is growing faster than the rest of the market, and at the current pace could reach half of all new releases as soon as 2027-2028. That is the conclusion of an analysis of roughly 53,600 games released on Valve's platform between July 2023 and July 2026, published by researcher Sulka Haro and covered by GRYOnline.pl. The same numbers show, however, that an individual game carrying an AI label performs noticeably worse on the market than a comparable title without one.

What the numbers show

The analysis covers every game released on Steam over three years, starting from the moment Valve became the first major digital distribution platform to require developers to explicitly disclose whether, and to what extent, they used generative AI in making a title. Right after the requirement took effect in February 2024, the number of AI-declared games jumped from a fraction of a percent to about 7 percent of all releases in that month, and it has climbed almost without interruption since.

According to Haro's calculations, 60 to 90 percent of the entire increase in new Steam releases in recent months has come from AI-labeled titles, while the number of releases without such a declaration has grown far more slowly, from about 1,030 to 1,320 per month. In other words, AI-assisted games are now driving most of the growth in new titles appearing on the platform each month.

Volume growth, not quality

The rising share of AI games in the platform's overall revenue does not mean individual titles are performing better than before. The author of the analysis stresses that this is purely an effect of scale, not an improvement in the quality of individual productions.

The growing AI share of sales is currently a story about volume, not quality - more attempts, not better conversion from individual AI games - Sulka Haro, author of the analysis

The data reveal a clear pattern among titles that actually make money. 72 percent of unsuccessful AI games limit their use of the technology mainly to graphics, textures and images. Among titles that achieved commercial success, AI is more often used to generate character voices (24 percent versus 8 percent among failures) and to localize games into other languages (18 percent versus 6 percent).

The mass of AI games that go nowhere is mostly low-effort, single-modality image generation - Sulka Haro, author of the analysis

Players increasingly check the labels

The growing number of AI-declared games is running into growing resistance from part of the Steam community. A separate study of 10,000 games on the platform, cited by GRYOnline.pl, found that titles labeled as made with AI receive on average 53 percent fewer reviews than comparable games without the label, which translates into weaker store visibility and lower buyer trust. Players have also started building their own tools to detect AI in games, including browser extensions that flag titles using generative models before purchase.

What it means for developers

For small studios and independent developers, who make up the majority of publishers on Steam, the analysis's conclusions are uncomfortable. Simply disclosing AI use guarantees neither higher sales nor a better reception, and according to the data, the median game with any reviews earns only about $300 across the entire platform, regardless of whether it used AI or not. 28 percent of all releases receive no reviews at all.

At the same time, the phenomenon is scaling up fast enough that, at the current pace, the share of AI-labeled games could exceed half of all Steam releases as early as 2027 or 2028. That is an extrapolation of the current trend rather than a hard forecast, but the direction of change is unmistakable: artificial intelligence is becoming a standard part of the game production process, regardless of whether it translates into commercial success for individual titles.

Polish context

For Polish game developers, who have long been one of the largest groups of independent publishers on Steam, the analysis's conclusions carry practical weight for production budget planning. The data suggest that simply reaching for generative AI for graphics or textures does not shorten the path to commercial success, and in some cases may even lower a title's credibility among buyers, who are reading labels more carefully and checking the origin of in-game assets.