Introducing the AI Disclosure & Scoring System (AIDSS): Replacing Binary AI Labels with Nuanced Transparency & Player-Controlled Filters

Summary:

Instead of a polarizing "all-or-nothing" AI ban or a single generic AI tag, we propose the AI Disclosure & Scoring System (AIDSS). This system allows creators to honestly disclose their specific AI usage across 7 categorized tiers during upload. Each tier contributes to an cumulative "AI Score" for the world/avatar. Players can then use a dynamic AI Threshold Filter in their settings to customize exactly what level of AI integration they are comfortable seeing in their search results.

Section 1: The 7 Tiers of AI Integration & Weight Scoring

We propose breaking down AI usage into seven distinct categories based on their level of impact on traditional human labor, ranging from mere productivity tools to fully automated generation:

Using DeepL or ChatGPT to translate world descriptions or menus into multiple languages.

Using GitHub Copilot or LLMs to help write, debug, or optimize specific snippets of Udon/C# scripts.

Utilizing Photoshop's Generative Fill or similar tools for seamless texture expansion, upscaling, or local image touch-ups.

Relying on AI tools for lightmap baking, auto-retopology, or reducing polygon counts to improve performance.

Core gameplay loops, whole prefabs, or complex interactive mechanics where the logic is almost entirely generated by AI code.

Worlds where 2D paintings, skyboxes, ambient music, or a significant portion of the material textures are direct outputs from AI generators (e.g., Midjourney, Stable Diffusion).

The layout, layout blocking, aesthetic style, and structural assembly of the world are entirely generated via one-click AI 3D space generation tools.

Section 2: UI Presentation (Creator Disclosure & Player View)

• Upon uploading, creators are presented with these 7 checkable categories.

• Voluntary Disclosure: Creators check the boxes that match their AI usage if they feel it reaches a meaningful threshold.

• Optional Description Notes: Next to each checked item, creators can type a brief note (e.g., "Only the skybox and a few wall posters were generated with Midjourney; the main architecture is entirely hand-modeled.")

• A clean "AI Disclosure Checklist" is displayed in the world/avatar details.

• Players can see the cumulative AI score and read the creator's descriptions, removing the "black box" anxiety and fostering mutual trust.

Section 3: Player-Controlled "AI Threshold Filter"

A sliding scale or numerical input will be added to the player's Account Gating Settings:

• Custom Gating Threshold (0 to 10 Points):

• 0 Points (The Pure Hand-Made Option): Content with any AI points (even a 1-pt translation help) will be completely hidden from their search and recommendations.

• 3 Points (The Moderate Option): Allows creators to utilize AI for coding help or image touch-ups, but automatically filters out low-effort "AI Slop" worlds that use heavy generative assets (3+ pts) or full world generators (10 pts).

• 10 Points (Unfiltered, the Experience-First Option): Default setting. Displays all content regardless of how it was made.

Section 4: Implementation & The "No Penalty" Policy (Anti-Cheating)

We intentionally propose not establishing a punitive or ban-based penalty system at launch.

• Why No Penalties?

VRChat currently lacks any objective, automated technological standard to reliably distinguish between AI-generated content and human-authored content. Implementing penalties would lead to false positives, endless appeals, and rampant witch-hunting.

• The Community-Driven Solution:

The system relies on honesty and crowdsourced transparency. However, once reliable detection tech or community evaluation standards mature, VRChat moderation staff could review reported worlds. If a creator is verified to have hidden their AI usage, moderators can manually force-check the appropriate AI categories on that world rather than banning the creator.

Why VRChat Team Should Adopt This System