Music partners should disclose the use of generative AI (Gen AI) in their content to help YouTube accurately treat and label media that contains Gen AI.
Disclose Gen AI content
You can provide information about Gen AI use when you deliver content to YouTube through DDEX or CSV delivery templates pursuant to the terms in your SRAV.
When you deliver your metadata, you can choose from the following affirmative designations:
- Fully Gen AI: All content was created using Generative AI.
- Partly Gen AI: Part of the content was created using Generative AI.
- No Gen AI: No content was created using Generative AI.
If you do not provide a value, YouTube will interpret it as unknown unless we are able to make a determination based on other signals.
It is up to partners to interpret the Gen AI definitions provided by DDEX in relation to their content. That said, here are some common scenarios and suggestions for how to classify content in those cases:
Full Gen AI:
- A user enters a text prompt into an AI music generator and downloads the resulting track.
- An artist uses a GenAI model to completely generate the composition, instrumental performance, and vocals for a song, using multiple prompts to iterate and refine the output to their liking before downloading the track.
Partly Gen AI:
- An artist uses a GenAI tool to generate one or more specific structural layers (e.g., a bassline or string section) and records their own live vocals and instruments over it.
- An artist uses GenAI tools to brainstorm lyrical themes, co-write verses, or generate and iterate on melodies, then records the audio in the studio.
- An artist produces a song using 100% human-performed audio, but generates the accompanying music video visuals using a text-to-video AI model.
No Gen AI:
- An engineer uses standard pitch-correction software (e.g., Auto-Tune) to correct minor tuning mistakes in a live vocal performance.
- A producer uses automated AI mastering software to adjust the final equalization, compression, and sonic balance of a track.