Valve Addresses Steam Next Fest AI and Low-Quality Game Discovery Problem

Valve has acknowledged a problem many Steam users and independent developers have been feeling during Steam Next Fest: when an enormous demo event begins by surfacing games somewhat randomly, low-quality projects and apparent clones can receive prominent exposure alongside games with far more distinctive ideas. Generative AI has sharpened that concern, not because copied concepts are new, but because the tools can make it easier to produce and publish more material at greater speed.

In remarks at Steam Indie Connect in South Korea, Steam Business Team members Tom Giardino and Marc Wilkinson described the issue as one Valve knows about, while making its underlying philosophy clear. The company does not want to dictate which development tools creators may use, including generative AI. Instead, Valve’s stated direction is to make Steam’s discovery system smarter and more dynamic without broadly restricting developer access.

That is not the same thing as saying the storefront’s current outcomes are ideal. Wilkinson specifically recognized that Steam Next Fest’s approach can expose low-quality games, and Giardino said there is ample room for improvement. The hard question is whether a better recommendation system, AI disclosures, and ongoing refinement can meaningfully reduce the visibility advantage enjoyed by disposable or derivative releases—without turning Steam into a platform that judges tools rather than finished work.

Why Steam Next Fest can look messy at the start

Steam Next Fest is a promotional event built around playable demos. For smaller teams, it is a chance to put an unfinished game in front of potential players, build wishlists, collect feedback, and test whether the pitch is connecting. But the event is no longer the comparatively small showcase Giardino recalled from earlier years, when developers needed encouragement to participate because there might be only 50 or 100 games.

That scale shift matters. A much larger submission pool creates a basic discovery problem: Steam cannot present every game to every player, so it has to decide what appears first and why. Wilkinson said the event begins with an element of randomness intended to make the system harder to abuse by people seeking immediate visibility. Later, more personalized recommendations take a greater role on the front page.

Personalized recommendations are suggestions tailored to an individual user rather than a single universal list. In a game store, that can mean the system increasingly uses signals connected to a player’s interests to decide what to show. The practical benefit is obvious: someone who plays strategy games should not need to sift through every genre to find relevant demos. The trade-off, as Valve’s comments make plain, is that the initial discovery period can place a broader and less consistently curated set of games in front of people.

The random opening is meant to support fairness and reduce manipulation, but fairness of opportunity does not automatically produce a satisfying first impression. If questionable projects reach high-visibility areas during those early days, they can shape how players perceive the whole event. They may also consume attention that a smaller, more carefully made demo desperately needs.

AI is part of the issue, not the entire issue

Giardino noted that clones predate generative AI. That distinction is important. A clone, in this context, is a game that closely imitates a familiar concept, mechanic, presentation, or commercially successful premise. Copying trends has been part of games for a long time, and banning a tool would not, by itself, eliminate derivative design.

Generative AI refers here to systems that can create or transform material such as text, images, audio, or other game-related content from prompts or inputs. Its relevance to Steam is not simply whether a title uses it. The larger concern is volume: if development and asset production become faster or cheaper for some publishers, more projects may be able to reach a store’s submission pipeline, including projects built around imitation rather than a clear creative identity.

Valve’s response does not frame AI use itself as an automatic quality failure. Giardino argued that developers should decide which tools to use rather than have the platform serve as a gatekeeper. At the same time, Steam requires disclosures for games that employ AI. That policy gives players information without declaring every use of the technology prohibited.

Disclosure is a meaningful baseline, but it is not a discovery solution on its own. A disclosure tells a potential customer something about a game’s production process. It does not determine whether that game appears on the front page, whether its demo is worthwhile, or whether it is crowding out projects that better match a player’s interests. The current dispute is therefore about both transparency and distribution: users want to know what they are seeing, but they also want confidence that the storefront is surfacing work worth their limited time.

Valve’s answer: better systems rather than blanket limits

Valve’s representatives presented a platform-level answer rather than a simple rule against AI-assisted games. Giardino said Steam should grow into a smarter, more dynamic platform, while acknowledging that Valve is not perfect. This puts the company’s emphasis on how the store sorts, recommends, and presents games instead of on a universal prohibition aimed at a category of production tools.

There is a coherent reason for that position. A rule targeting a tool can be difficult to draw and enforce consistently. “Uses AI” can cover very different circumstances, from a limited element of production to a game whose public-facing material is heavily generated. More importantly, a tool-focused rule does not directly answer the player’s practical concern: whether the store is highlighting something original, clear, appealing, and relevant to them.

But the approach also places a large burden on the storefront’s systems. If Valve wants to preserve broad access for creators, those systems must become good enough to distinguish meaningful player interest from opportunistic attempts to seize exposure. The company has not provided a specific new fix in these remarks, only confirmation that many people are working on the problem.

That leaves the central tension unresolved. The random discovery layer is described as a defense against visibility abuse, yet it can provide early visibility to games players regard as poor or derivative. Personalization can improve relevance later, yet it may not undo the early impression or attention already created. A smarter system will need to handle both sides: resist manipulation while still helping genuinely distinctive demos find their audiences.

What “success” means is different for every developer

One useful part of Giardino’s comments is the reminder that Steam Next Fest does not have one universal success metric. For one team, 1,000 wishlists might be a meaningful result. For another, the target could be 50,000. A wishlist is a user’s indication that they may want to buy or follow a game later, so it is a valuable expression of interest—but it is not a single, fixed scorecard for every project.

That makes platform responsibility more complicated. Valve cannot sensibly promise that every game will reach the same number of players or generate the same wishlist total. Nor can it define what a successful event looks like for every developer. It can, however, work on the quality of the route between a game and the people most likely to care about it.

For a small studio, that distinction is not academic. A short period of visibility can influence feedback, community growth, and whether prospective players even learn the demo exists. In a crowded event, originality alone may not guarantee attention. Giardino’s advice was that games should continue adding features, personality, and creativity rooted in their creators’ own perspective rather than relying on one mechanic that can be copied easily.

That is strong creative advice, although it does not remove the structural challenge. A developer can make a more personal game and still face a storefront where hundreds or thousands of other projects compete for a fleeting glance. The value of a recommendation system is not that it declares which game deserves to win; it is that it can give more people a plausible path to the games they are actually looking for.

What players should watch for next

The immediate takeaway is not that Valve has announced a ban, a new review program, or a specific change to Next Fest’s opening recommendations. It has not. The acknowledgement is narrower but still significant: Valve recognizes that low-quality games can be exposed by the event’s current trade-offs, recognizes the scale of the problem, and is considering how to address it while maintaining an open approach to developer tools.

For players, Steam’s AI disclosure requirement remains the clearest existing safeguard mentioned here. It offers a piece of context when deciding which demos to download or which games to wishlist. For developers, the discussion reinforces that event participation takes place inside a highly competitive discovery environment, where a distinctive voice and a clear game identity may matter as much as matching a popular mechanical trend.

The broader industry picture is one of uneven visibility rather than a simple AI-versus-non-AI divide. Large audiences and strong demand can look very different from project to project; even estimates around major releases such as Phantom Blade Zero’s reported pre-launch sales underline how far market outcomes can sit from the goals of a typical demo participant. Steam Next Fest’s challenge is to remain useful for the smaller game seeking its first audience while serving players who do not want to dig through noise to find it.

Valve’s stated preference is to solve that challenge through better discovery rather than sweeping limits on generative AI. Whether that approach satisfies Steam users will depend on visible improvements: fewer weak or imitative projects taking over initial discovery, more relevant recommendations, and clearer evidence that a creator with a compelling demo can still be found in an event that has grown far beyond its early scale.