Social VR platform VRChat has decided not to implement “AI-generated” content labels or user filters to exclude AI content. Instead, it will prioritize developing discovery features that allow users to directly rate world quality and view multiple world images before entering.
| 📒 | – VRChat will not introduce AI-generated labels or exclusion filters – Disagreements over labeling criteria and concerns over creator harassment – World ratings targeted for late 2026, galleries for H1 2027 |
VRChat announced the decision on its official forum on October 28 while revealing the results of a week-long survey filled out by 21,595 users. Although 95% of respondents wanted some form of label indicating AI usage, answers split across 3,141 different combinations when asked which specific items among eight example categories should be labeled, with no single combination exceeding 10% of respondents.
Wanting labels did not necessarily mean users intended to avoid AI content. Only 33% of respondents said they would intentionally avoid “AI-assisted” worlds. Furthermore, 37% of users who submitted written feedback expressed that using AI as a tool is acceptable as long as a human leads the creative process. VRChat added that some long-time creators have hesitated to share their work out of fear of harassment if their use of AI were disclosed.
Synthesizing survey results and one-on-one creator interviews, VRChat concluded that relying on self-reported creator labels could not guarantee accuracy. Creators could try to pass off generative or assisted work as entirely crafted by hand, while a culture of “sleuthing” out AI usage could spiral into harassment targeting creators who did not use AI tools at all. The platform also noted that exclusion filters rely on the premise of accurate labeling and warned that, with the definition of AI constantly shifting, setting a filter once could cause users to permanently miss out on certain content.
Instead, VRChat will first revamp ratings and informational features to help users evaluate worlds. It plans to add a “Likes/Dislikes” rating system to worlds, aiming for release by late 2026 or early 2027 if delayed. Once enough data is collected, the company will explore how to display ratings and integrate them into recommendations. To mitigate issues like vote manipulation or low-effort, attention-grabbing worlds gaining an unfair advantage, the platform also plans to factor in evaluator activity levels.
Following the rating system, VRChat aims to release multi-image galleries in the first half of 2027, expanding beyond the current format that displays only a single thumbnail before joining a world. VRChat previously updated its Creator Guidelines to prohibit misleading thumbnails. The upcoming SDK update will also include a notification reminding creators of these rules when selecting a world thumbnail. Target timelines for both world ratings and galleries remain subject to change.
Under consideration are an optional “Human-authored” mark for worlds built without generative AI and a “Tools used” field where creators can list the production tools they used. It remains undecided whether the standard for “Human-authored” requires completely excluding AI tools or simply having a human lead the creative process. Neither feature has been confirmed for implementation, nor has a release timeline been set, though VRChat stated it will share an update by the end of Q1 2027. Displaying videos and panoramas, alongside world reviews and comments, have also been placed on the list for future consideration.
Originally written in Korean, this article was translated with AI assistance and edited by a native English-speaking editor. Every translation is reviewed by our editorial team before publication.
[Read Original]
