YouTube Music began labelling songs made primarily with generative AI on 14 October, with a tag on the player page whenever “AI played a significant role” in a track, and the labels will later reach the main YouTube app.
The example YouTube gave in its announcement is a track where a generative model created “the composition and vocals.” A day earlier, Deezer said it would let listeners remove fully AI-generated music from search altogether.
How YouTube decides what gets a label
YouTube says it is “teaming up with music partners” and using its own detection tools to find AI tracks. Tapping the tag opens a panel headed “How this was made”, which reads: “This audio was altered or fully generated using AI,” according to 9to5Google.
Labels and distributors declare AI use when they deliver music, through DDEX metadata, a CSV template or YouTube Studio. The YouTube Help page for music partners gives three choices: Fully Gen AI, Partly Gen AI and No Gen AI.
A song typed into an AI generator as a text prompt counts as fully AI. So does a track whose composition, instrumental performance and vocals all came from a model, even after many rounds of prompting.
Partly AI covers an artist who generates one layer, such as a bassline or a string part, and records their own vocals and instruments over it. Using AI to brainstorm lyrics or melodies before recording in a studio also counts, as does a human-played song with an AI-made music video.
Fixing a vocal with Auto-Tune, or letting automated AI mastering software set a track’s final EQ and compression, is not generative AI under YouTube’s guidance.
If a partner says nothing, YouTube “may designate content as fully or partly Gen AI” from its own signals.
Deezer says it has caught 42 million AI tracks
Deezer’s toggle, announced in Paris on 13 October, will hide AI-generated tracks, albums and the artists behind them from search results. It will roll out worldwide before the end of 2026, and users can switch it off again.
The company says it found more than 42 million fully AI-generated tracks between January 2025 and August 2026 and kept them out of its recommendations. It has tagged over 14 million albums since it began labelling AI music in June 2025.
Deezer now receives nearly 90,000 fully AI-generated tracks a day, more than half of all deliveries on peak days, according to its newsroom.
Those tracks get only 1% to 3% of streams on the service. Deezer found that up to 85% of the streams they did get in 2025 were fraudulent, and it pays no royalties on manipulated plays.
“We’ve seen a clear user demand for the ability to filter out AI-generated music,” said Alexis Lanternier, Deezer’s chief executive. A Deezer/Ipsos survey from November 2025 found that 45% of streaming users wanted to filter out music made entirely by AI.
Spotify, Apple Music and Tidal have their own rules
Spotify brought in a voluntary AI tag earlier this year and said it would not recommend songs from “AI personas”. Apple Music said in August that it would add AI labels to songs later this year, Variety reported.
Tidal labels recordings it identifies as entirely AI-generated, lets listeners turn them off and pays no royalties on them, according to TechCrunch.
Over the summer the RIAA, the IFPI, the Recording Academy and SAG-AFTRA asked services to mark songs as “AI-Generated” or “AI-Assisted”. “Fans want to know whether and how generative AI has been used in the music to which they listen,” the heads of the IFPI and the RIAA said.
Fake streams are already a criminal matter: one man got 18 months for AI streaming fraud. DistroKid has been pulling songs over a Universal lawsuit, and Italy is investigating Suno, the app that makes songs from text prompts. Small labels should also know Spotify’s 1,000-stream rule.
YouTube also runs its own AI music tool, Dream Track, which lets selected creators make soundtracks with AI versions of the voices of artists who have signed up.
Lead photo: Mihis Alex / Pexels, Pexels License