Does YouTube make you disclose AI music? The label's rules, and the two gaps in them

What happened

YouTube began rolling out a "Made with AI" label on October 14, 2026, according to The Verge. It marks tracks that were fully generated or substantially altered with generative AI — including AI-written composition and AI vocals.

The label is showing up in YouTube Music first. YouTube says it is expected to expand to the main YouTube app.

Two details matter more than the label itself. First, music partners are expected to declare AI use at upload. Second, if they don't, YouTube says it can use its own signals to decide a track is AI-generated anyway. That is a disclosure duty with a detection backstop behind it.

What counts, and what the reporting does not establish

In disclosure territory, per YouTube's description of the system: AI-generated composition, AI-generated vocals, tracks substantially altered by generative models, and synthetic versions of recognizable performers.

Not established as automatically covered: standard pitch correction, ordinary editing, mixing and mastering, or AI used only on scripts and metadata around a track. YouTube's stated target is generative AI producing substantial musical elements — not every tool in a DAW.

Here is what nobody has published. There is no threshold for how much generative AI makes a track "substantially altered." There is no number for how many tracks will get the label. There is no stated figure for how often detection signals override what a partner declared. If you see one quoted this week, ask where it came from.

What this is NOT

This is not a demonetization rule. The confirmed change is identification and disclosure. The available reporting does not establish that a labeled track loses monetization, loses recommendations, or gets downranked. Treat "labeled" and "penalized" as separate questions until YouTube says otherwise.

It is also not a clearance. This is the part that burns video publishers. You can hold a valid license for a song and still owe YouTube a disclosure that the song was AI-generated. Rights and transparency run on different tracks. Being legally allowed to use it says nothing about whether it has to be labeled.

And it is not the same policy everywhere. Spotify introduced a voluntary AI label and said it would not recommend songs from "AI personas." Apple Music was reported to be considering an AI tag. Spotify's version carries a recommendation consequence that YouTube has not announced. Do not assume one platform's rule describes the others.

Why it matters if you publish

The exposure is not in music channels. It is in everyone who scores video with cheap library tracks.

A large share of royalty-free catalogs added generative output in the last two years, often without flagging it downstream. If the label migrates to the main YouTube app as expected, your back catalog of uploads inherits whatever determination YouTube makes about the music in them. You will not be asked first.

The practical problem is that most creators cannot answer the question "was this track AI-generated?" about music they licensed 14 months ago. The supplier knows. You don't. That gap is the work.

What to do about it

1. List every music source you used in the last 12 months — library subscription, marketplace, commissioned composer, AI generator.
2. Email each supplier one question: was any track in my downloads generated or substantially altered by generative AI? Keep the reply.
3. If you generate music yourself, disclose it at upload rather than waiting for detection. The backstop exists; being the one who declared is a better position than being the one who was caught.
4. Keep license files and AI-origin answers in the same folder. They are two separate claims and you will eventually need both.
5. Do not rebuild your soundtrack strategy on an assumed penalty. No monetization or recommendation consequence has been announced for YouTube's label.
6. Re-check YouTube's music policy page when the label reaches the main app. The thresholds are unpublished today; the first specific number is the thing that changes your workflow.