YouTube Music Introduces Labels for AI-Generated Songs
YouTube Music announced on Wednesday, 14 October, that it was rolling out labels identifying songs created fully or partly with generative artificial intelligence (AI). The announcement did not specify when the rollout began. The labels appear on the player page, and selecting one opens an explanatory panel describing how AI was used to make the track. The categories cover songs made entirely by AI and tracks where AI contributed to the composition or vocals.
The system has three designations: Fully Gen AI, Partly Gen AI and No Gen AI. Music labels and distributors are expected to disclose generative AI use so YouTube can identify and label the content appropriately.
Fully Gen AI applies to music created entirely through generative AI. This includes tracks produced when a user enters a text prompt into a music generator and downloads the result. It also covers recordings in which a model generates the composition, instrumental performance and vocals.
Partly Gen AI applies when AI contributes to some elements while people create or perform others. An artist might use AI to produce a bassline or string section before recording live vocals and instruments. The category also covers using AI to co-write verses or brainstorm lyrical themes.
No Gen AI covers production techniques that do not qualify as generative AI under the labelling system. Conventional pitch correction, including Auto-Tune, falls into this category, as does automated AI mastering. These examples distinguish the production techniques covered by the category from the generation of musical content.
YouTube also uses its own detection tools to identify AI-generated music. If a music partner fails to disclose such use, the platform may rely on its own signals and label a track as fully or partly AI-generated.
YouTube has not disclosed the accuracy of its AI detection tools or the proportion of its music catalogue containing AI-generated material. Its announcement also does not clarify whether Partly Gen AI labels will appear on the player page in every case.
AI involvement alone does not make a track ineligible for monetisation. Fully AI-generated songs can still earn revenue, and the labels are not intended to change recommendations automatically. YouTube Music does not offer a filter to hide AI-generated tracks, so listeners can identify them but cannot exclude them through a dedicated setting.
Other streaming services have adopted different policies. In July, Deezer reported receiving nearly 90,000 fully AI-generated tracks each day. The service labels detected tracks and excludes them from algorithmic recommendations and editorial playlists. It also withholds royalties from streams it deems fraudulent.
TIDAL labels recordings it identifies as fully AI-generated. It allows listeners to switch off these tracks and excludes them from royalty attribution. YouTube Music does not automatically restrict recommendations or monetisation solely because a track was made with AI.
YouTube also runs Dream Track, an AI music experiment that lets selected creators produce soundtracks using AI versions of participating artists’ voices. The platform’s labelling system identifies AI involvement in music, while Dream Track allows selected creators to use AI-generated versions of participating artists’ voices in their work.
The company has previously faced a dispute over synthetic vocals. The viral track Heart On My Sleeve featured AI-generated vocals imitating The Weeknd and Drake. Universal Music Group secured its removal. The incident preceded YouTube’s introduction of requirements for creators to label realistic AI-generated content.
