Human-made, AI-assisted, hybrid: why we label our work

Every image and film we publish carries one label. It tells the viewer how the piece was made, so they do not have to guess. There are three labels, and each has a plain meaning.

Human-made means people made the piece from start to finish. AI-assisted means people led the work and used AI tools for specific tasks, and people reviewed every result. Hybrid means human and AI work were combined into one piece, directed and approved by people.

Why three and not two? A yes-or-no label would put a retouched photograph and a generated image in the same box. Both used AI, but they are not the same kind of work. Three labels keep them apart, and tell the viewer how much of what they see a person made.

We expect three labels to be clearer to viewers than a yes-or-no one. We have not tested that yet. In early conversations, people asked which label a piece carried before they asked who made it.

Most pieces are easy to label. The hard cases sit at the edges, and we settle them with two questions.

First, did an AI tool do any task in making the piece? If not, the piece is Human-made. Cleaning up sound, removing a background and matching color all count as tasks. Ordinary tools that follow a person's hand, like a camera or editing software used by hand, do not.

Second, did AI make anything new that the viewer sees or hears? If it did, the piece is Hybrid. Generated backgrounds, voices and faces all count. If AI only did tasks and made nothing new, the piece is AI-assisted.

Some cases still need judgment. When a piece is hard to call, our creative direction lead decides and writes down why. Those reasons become part of this article.

There is no fourth label for work made by AI alone. We do not publish work that no person directed.

A label is not a grade. It says how a piece was made, not whether it is good. A Human-made piece is not better because of its label. We want the work judged as work.

You can use the same labels on your own work today. Ask the two questions before a piece goes out, and put the label next to the credits. Keep a short note of how you decided. It matters when someone asks.

This article will grow as new hard cases come up. When we test the labels with viewers, we will publish what we find, including where the labels confused people.