VRChat Rejects AI Labels and Filters as Steam Reviews Turn Mostly Negative

VRChat will not add an AI-generated label or AI filters to its user-made worlds, even though 95% of the 21,595 people who answered its survey wanted some form of AI disclosure, Inven Global reported.

The social VR platform announced the decision on its official forum on 28 October.

Players have not taken it well. mixed-news.com reported that VRChat's recent Steam reviews carried a Mostly Negative rating on 2 November, at 32% positive across 4,155 reviews from the past 30 days.

That figure covers the whole period and does not show how many reviews were about the AI decision.

Players Push Back

The same outlet noted that a public feature request on VRChat's feedback board, asking for AI disclosure and filtering, has attracted more than 1,000 votes. Some commenters have criticised the company's response.

The survey reflects only the people who chose to answer it.

Why VRChat Rejected Labels and Filters

VRChat's case against labels rests on self-reporting. Creators could present generative or AI-assisted work as fully hand-made, and a culture of hunting for AI use could lead to harassment of creators who never used it.

VRChat said that risked "negative behavior spirals we don’t want to encourage," shanethegamer.com (Shane the Gamer) reported.

Filters were ruled out separately. VRChat argues a player switches one on once and then permanently misses content, and that a filter depends on accurate labels. It also says the definition of AI keeps shifting.

Shane the Gamer pointed to a practical obstacle as well. VRChat has found its ban on misleading world thumbnails hard to enforce because it has no automated way to spot a misleading image. Checking claims about how a whole world was built would be harder.

What the Survey Found

Support for a label did not mean support for avoiding AI work. Asked which of eight example categories of AI use should be labelled, respondents produced 3,141 different combinations, and none was picked by more than 10%.

VRChat says the figures are weighted to reflect its player base and that results for smaller groups are less certain.

Only 33% said they would intentionally avoid AI-assisted worlds, according to Inven Global. Among users who left written feedback, 37% said AI is acceptable as a tool as long as a human leads the creative process.

VRChat added that some long-time creators have held back from sharing work for fear of harassment if they disclosed AI use.

The least divisive ideas were guidelines against misleading thumbnails (81% rated them very useful), discovery that rewards quality (72%) and in-game feedback (69%), Shane the Gamer reported.

What VRChat Is Building Instead

VRChat is adding a Likes/Dislikes rating system for worlds, targeted for late 2026 or early 2027 if it slips. Once enough data is in, the company will work out how to display ratings and feed them into recommendations.

It also plans to factor in how active each rater is, to limit vote manipulation and low-effort worlds gaining an unfair edge.

Multi-image galleries are due in the first half of 2027. Worlds currently show a single thumbnail before players join. An upcoming SDK update will also remind creators of the thumbnail rules when they pick an image. VRChat describes the timelines as estimates.

Two transparency features remain under consideration, with no commitment. One is an optional Human-authored mark, which mixed-news.com reported would not affect discovery rankings. The other is a Tools used field where creators list what they built with.

VRChat has not decided whether Human-authored would mean no AI tools at all or simply a human-led process. The company says it will share an update by the end of the first quarter of 2027.

The post VRChat Rejects AI Labels and Filters as Steam Reviews Turn Mostly Negative appeared first on Stevivor.