VRChat Refused to Label AI Content. Here's the Distinction That Debate Is Missing.

VRChat is currently weathering a sizable backlash after deciding not to label AI-generated content on its platform.

The sequence is worth recapping precisely. Earlier this month, VRChat released a survey asking users whether the platform should in some way label AI-generated content — explicitly framed as a way to fight back against low-quality "slop." Users answered. Then the platform declined to move forward with labeling. The community response has been loud, and it's not hard to understand why: people were asked their opinion about the thing degrading their experience, and then told no.

That's a legitimate grievance. But somewhere inside it is a distinction that most of the commentary is flattening, and it matters for anyone building or buying VR software right now.

When people say AI in VR, they usually mean one of two very different things.

Generative AI that produces content. Textures, meshes, avatars, worlds, voices. It scales output without scaling authorship, and when it's unlabeled, the user has no way to weigh what they're looking at. This is what the VRChat survey was about. The backlash is earned.

Deterministic AI that operates as an instrument. It doesn't create anything. It sits between a person and content they already chose, and modifies how that content is perceived. Its value is measurable in whether a task gets easier, not in whether the volume goes up.

The labeling debate treats these as the same policy question. They aren't.

Amblyotube, developed by Seven Sports, is a Meta Quest app built for people with amblyopia — lazy eye — and for anyone working on visual coordination and attention. It's a clean example of the second category, because it generates nothing at all.

The premise: you watch YouTube-style video, the kind of content you'd watch anyway, and the app renders a different visual experience to each eye.

No adhesive patches. No red-blue glasses. A session runs 30–40 minutes, done with content you enjoy, inside a headset you may already own.

If the VRChat episode produces a policy, it should be a policy about disclosure of generation, not about the presence of the word AI.

The user's actual complaint is: I can't tell what was made by a machine, and nobody will tell me. That's a transparency failure. The remedy is labeling generated assets.

A different question is: does the software use computation to change what I perceive, and is that stated plainly? For an assistive or training tool, that's not a hidden cost — that's the entire product. The honest move isn't a generic AI badge; it's a clear statement of what the processing does and what it is not.

Which is why the boundaries of Amblyotube are stated as flatly as its features:

That last line is not marketing hedging. Amblyopia care involves professionals. Software like this is something you do alongside that, not instead of it.

VRChat's users aren't angry that AI exists in their space. They're angry that its presence is being concealed. Those are different problems with different fixes, and conflating them produces bad rules that punish transparent instruments while letting unlabeled generation slide through.

Ask what the AI is doing. Ask who it's accountable to. Then label accordingly.

Amblyotube is available on the Meta Quest app store: https://www.meta.com/en-gb/experiences/amblyotube/25906906972338493/