Three labels, one bright line

Partners declare generative AI use when they deliver tracks to YouTube via DDEX or CSV templates, choosing from three designations. “Fully Gen AI” covers the obvious case: a track downloaded from an AI music generator after a text prompt — composition and vocals made by the model, as Music Business Worldwide reports. “Partly Gen AI” is the murky middle: an artist who generates a bassline or string section with AI, then adds live vocals and instruments on top. YouTube’s own examples stretch the category further — a song with 100% human-performed audio still qualifies as “Partly Gen AI” if its music video was made with a text-to-video model.

“No Gen AI” is the carve-out: pitch-correction software like Auto-Tune fixing minor vocal mistakes, and automated AI mastering that sets the final balance, compression and EQ, both land in the no-label bucket, per YouTube’s help page. The line isn’t whether AI touched the track — it’s whether AI generated the creative substance.

Music producer working at a computer with mixing equipment in a dark studio
A producer’s desk, human or otherwise: YouTube’s three-tier labels distinguish full AI generation from AI-assisted tracks. Stock photo: Pexels.

The detection net

The system leans on partner disclosures, but YouTube isn’t taking everyone’s word for it. It says it will use “additional signals to augment partner declarations” — its own detection tools — and where partners provide no Gen AI information at all, YouTube will rely solely on those signals and “may designate content as fully or partly Gen AI.” Disclosure is encouraged; the platform can label you anyway.

One thing YouTube didn’t say: whether the “Partly Gen AI” tier gets the player-page label at all. The announcement promises the label for tracks where AI played a “significant role,” but the help-center examples suggest the partly tier is a delivery-time declaration that may or may not surface to listeners, TechCrunch notes.

Audio engineer editing sound on a computer in a recording studio
Editing the mix: YouTube will combine partner disclosures with its own AI-detection signals. Stock photo: Pexels.

Transparency, not treatment

Crucially, neither the blog post nor the help page says what happens to labeled tracks — no word on whether they’ll be recommended differently or paid differently. The post does link back to YouTube’s May update on AI video labels, where the company said plainly that “a disclosure label alone does not change how a video is recommended or whether it’s eligible to earn money.”

That’s a gentler stance than some rivals are taking. TIDAL began tagging fully AI-generated music in June and went further, stopping royalty payments on it. Spotify has been testing AI disclosures in song credits since April and in August added an “AI Persona” badge for AI-generated artist identities, keeping such music out of editorial and algorithmic recommendations by default. Apple Music told industry partners in August that “Made With AI” tags would reach listeners later this year. And Deezer — which says it receives nearly 90,000 fully AI-generated tracks every day — announced Tuesday it will let users exclude fully AI-generated music from search results.

Why it matters

The scale is the story. Tens of thousands of fully AI-generated tracks are flooding streaming services daily, and the industry spent the summer groping for a labeling vocabulary — the RIAA and IFPI even proposed two voluntary labels in July, “AI-generated” and “AI-assisted.” YouTube, the world’s biggest music platform, has now picked its side: label loudly, but keep the songs earning. As the company put it, “our listeners have told us they want transparency when they’re hearing an AI-generated track.” The label tells you a song was made with AI; it doesn’t judge it, throttle it, or starve it. Whether listeners change their behavior when they can see the tag is the experiment everyone will be watching.