Why Steam Can’t Ban AI Slop Games: And What It Can Do Instead

Your anger at games cranked out entirely by AI is understandable. One developer analyzed 53,597 Steam games and found that AI games were only 55% as likely as non-AI games to receive more than 100 reviews. That gap had not narrowed in two years. About 87% of AI games sold poorly. In one week, Steam released more than 720 games; nearly 75% had fewer than 10 reviews, and 530 had only 0 to 9. Many games are cannon fodder from the day they launch.

But “the market is eliminating AI slop” and “Steam can completely block AI slop” are different claims. The first is happening. The second is almost impossible to execute. The problem is not that the platform is lazy. The problem is that a full ban fails at three levels: definition, detection, and platform logic.

Market screening works, but it eliminates games, not behavior

Your cost estimate is roughly right. Steam charges $100 to list a game. AI-generated art and code cost a few hundred yuan. Add time and other expenses, and a bare-bones release can cost around one thousand yuan. If nobody buys it, and even recouping costs is hard, AI slop should decrease.

That logic holds for one developer. It breaks down for mass speculation. Speculators are not betting that every game succeeds. They are betting that if they cast a wide net, one hit will cover the rest. For a serious developer, the listing fee is a barrier. For a mass speculator, it is a trial-and-error cost spread across many releases. Massive releases also dilute exposure. When 720 games launch in a week, players cannot give each one attention. For games nobody notices, whether they used AI probably makes no difference to whether anyone notices them.

So market screening lets most AI slop sink to the bottom. It does not make the behavior disappear. A full ban would require the platform to judge, before the market has tested anything, which games are doomed to be junk. That is almost impossible.

“Whom would you block?” is an unanswerable question

You are objecting to shoddy production. But AI involvement does not equal shoddy production. If the platform wants to block something, the first question is: whom to block?

Steam’s current AI disclosure rules are selective. Under the rules adjusted in January 2026, AI-assisted coding and the use of AI tools to improve efficiency do not need to be disclosed to players. Only pre-generated content visible to players, such as art, audio, and text, and live-generated content inside games, must be disclosed. The policy is still considered voluntary in practice and is not enforced through mandatory review by Valve.

That creates a loophole. A game can use AI heavily to generate code logic. If the art and text are human-made, it can legally carry no AI label. A carefully made game that uses AI for part of its art and discloses it honestly will be labeled.

A full ban would require a detection system that can penetrate the code level and measure “AI content.” That is technically infeasible, and it would hurt many normal developers who use AI as an efficiency tool. Nobody today labels a game “not purely handmade” because its developer used code-completion tools. Epic’s CEO criticized Steam’s disclosure policy as “irresponsible.” The core argument is that AI tools will become ubiquitous, and labeling “use of AI” separately will eventually make the label meaningless.

Steam already withdrew from the referee role

Steam does not easily take a one-size-fits-all approach to content. In 2017, it replaced the Greenlight community-voting system with Steam Direct, which lets a developer list a game for $100. The change lowered the threshold for independent developers and shifted the judgment of “what deserves to exist” from platform review to market feedback and algorithmic distribution.

A preemptive ban on “AI slop” would ask Steam to abandon the open-distribution logic it has relied on for nearly a decade. It would also ask the platform to return to the role of content referee that it voluntarily gave up. Steam is more likely to mandate disclosure of core AI content, add player reporting, downrank algorithmically, and provide filtering tools. It is less likely to decide for every player which game does not deserve to exist.

The cost of junk lands on players, not the platform

Blocking has another problem. The cost of junk content is mainly borne by players and by normal developers who get drowned out. Massive releases dilute exposure. Players have limited attention. Many mid-tier games that are actually decent get buried. That is the externality you described: junk consumes attention and pushes better games out of view.

Steam’s $100 listing fee is both a barrier and a stable source of platform revenue unrelated to quality. Asking the platform to cut off that revenue in order to carry out a task that is technically vague, philosophically contradictory, and costly mainly to others is not realistic. The lack of incentive is structural, not moral.

A better path: disclosure, filtering, and player choice

Your desire to keep AI slop out of sight is reasonable. The closest practical option is not a platform ban. It is to let players filter for themselves. The third-party database SteamDB already offers a filter based on Steam’s disclosure data. Players can exclude games that have disclosed the use of generative AI. This turns “blocking” from platform-level content review into player-level preference.

This path is imperfect. It depends on honest disclosure. It may also hurt games that label their AI use honestly but are still decent. But it avoids the definition problems and collateral damage of a blanket ban, and it leaves the choice with players. Steam’s current approach, mandatory disclosure of core AI content plus a player reporting mechanism, follows the same logic.

A more realistic governance mix would include clearer disclosure rules, easier filtering tools, player reports, algorithmic downranking, and penalties for clearly deceptive marketing. The platform does not need to decide what “AI slop” is. It can make “whether AI was used” and “where AI was used” transparent, then let players decide whether to buy.

Conclusion

The disgust is real. It comes from low-cost speculation capturing public attention. But a complete block cannot work. Under the current platform architecture, shoddy production cannot be identified and intercepted by an automated rule before publication. After-the-fact market punishment is not enough to make the behavior disappear at its source.

Steam is not unable to govern. It is unable to govern by complete blocking. It can improve transparency, lower the cost of filtering, and punish deceptive marketing. Players can vote with their wallets and use filters to protect their attention. The market can let most AI slop sink. How far AI can actually take game development needs time to test. Players are not disgusted by the tool itself. They are disgusted by people who enter the market with speculative motives and produce shoddy results. Those people have stigmatized the “AI game” label.

Instead of expecting Steam to become a firewall, push it to become a clearer label and a better filter. In an open market, a platform can hardly eliminate low-quality content. It can make that content less visible, and players can push its visibility lower.