YouTube Music now labels AI songs: what 'Fully', 'Partly' and 'No Gen AI' mean for you

YouTube Music began labelling songs "created primarily using generative AI" on 14 Oct 2026. The label sits on the player page, tapping it opens more detail, and YouTube says the labels "will also expand to the main YouTube app". It identifies tracks by working with music partners and running its own detection tools.

We covered how this fits next to Apple, Spotify, TIDAL and Deezer on launch day. This piece is the practical one: what each label means, how a release gets one, what it does to your money, and how it differs from Spotify's AI Credits.

What do the three labels mean?

YouTube's help page for music partners defines three declarations and gives examples for each:

• Fully Gen AI: all of the content was made with generative AI. Examples: typing a prompt into an AI music generator and downloading the track, or having AI generate the composition, instrumental and vocals.
• Partly Gen AI: some of it was. Examples: generating a bassline or string section and adding your own vocals and instruments; using AI tools to brainstorm lyrical themes, co-write verses or generate and iterate on melodies; or a song with fully human-performed audio and an AI-generated music video.
• No Gen AI: none of it was. The examples are an engineer using standard pitch-correction software such as Auto-Tune, and a producer using automated AI mastering to set EQ, compression and balance.

Two lines in that list deserve attention. First, songwriting counts. If you used a chatbot to rough out a verse or a melody generator to find a hook, YouTube's examples put the release in Partly Gen AI, even if every note was then sung and played by people. Second, "AI" in a tool's marketing doesn't decide the label. AI mastering is No Gen AI, because it processes your audio rather than generating new material. TechCrunch puts it simply: pitch-corrected tracks fall under No Gen AI.

How do you declare?

You probably don't do it yourself. Declarations come from music partners, which for most independent artists means your distributor. The help page lists three routes:

1. DDEX delivery, ERN 4.3.2 or later, under the partner's agreement with YouTube.
2. A CSV template in YouTube Studio Content Manager, with a contains_ai column that takes all, partly, none, or unknown to clear an earlier value.
3. YouTube Studio's attribute section, for partners who don't deliver through DDEX or CSV.

The help page also leaves it "up to partners" to interpret the DDEX definitions for their own content. So the practical steps for an artist are:

• Ask your distributor whether it passes a Gen AI declaration to YouTube, and where in its upload form you set it.
• Decide your category per release, using YouTube's examples rather than your gut. AI lyric help is Partly; Auto-Tune is not.
• Write one sentence recording what AI did on the release, and keep it with your project files. If a label looks wrong later, that note is your evidence.
• Fix wrong values early. The CSV unknown value exists to clear a previous declaration, so ask your distributor to correct it rather than waiting.

What happens if nobody declares?

According to the help page, a missing value is read as unknown, and YouTube "may use additional signals to augment partner declarations". With no partner declaration, it relies on those signals alone and may label a track fully or partly Gen AI. Neither page says how accurate the detection is or how to dispute a label. Music Business Worldwide points to a different gap: neither page says whether labelled tracks will be handled differently in recommendations or royalty payouts. It is also unclear whether Partly Gen AI tracks show a visible label, since the blog post describes labels for songs made "primarily" with AI.

Does a label affect what you earn?

TechCrunch reports that YouTube is favouring disclosure over restriction, and that AI involvement alone doesn't make a track ineligible for monetisation. That is TechCrunch's description; YouTube's blog post and help page don't mention money or recommendations at all.

The rule that can cost money on YouTube is a different one. Its channel monetisation policies for the Partner Program rename "repetitious content" as "inauthentic content" and exclude AI-generated content made with generic templates that gives the impression of mass production. That applies to channels, and it judges output, not the label.

How does this compare with Spotify's AI Credits?

Spotify launched AI Credits in April and says artists submit tens of thousands a day, each showing which roles and instruments used AI. The two systems differ in ways that matter:

• Granularity. Spotify's AI Credits cover lyrics, vocals, instrumentals and production separately. YouTube gives one designation per asset.
• Visibility. Spotify's credits sit in the song credits on mobile during the beta, not as a track label. YouTube's label is on the player.
• Silence. Spotify says outright that the absence of AI credits doesn't mean AI wasn't used. YouTube may fill silence with its own call.
• Access. Spotify's credits are optional and, during the beta, only available with certain distributors. YouTube's help page describes declarations through DDEX, CSV or YouTube Studio, with no beta restriction mentioned.

For how Spotify treats tracks it judges to be "slop", which is a separate system, see our piece on its 50 million recommendation removals.

What we think

YouTube has drawn the assistive-versus-generative line in roughly the right place, and then blurred it. Treating Auto-Tune and AI mastering as No Gen AI is correct: they process a performance rather than invent one. But one Partly bucket holding both an AI-generated string section and a chorus idea you got from a chatbot and then rewrote tells listeners almost nothing. Spotify's per-role credits are the better model for honest artists, because "AI on lyrics" and "AI on vocals" are very different claims.

The real shift is that silence now has a cost. In Spotify's AI Credits, not declaring just leaves a blank. On YouTube, our reading of the help page is that the detector decides for you, with no published appeal route. The artists most exposed are the ones in the middle: people who write and perform their music but use generative tools for a layer or a sketch. That covers AI co-producers such as Veena, which builds tracks with you layer by layer: the honest declaration there is usually Partly, and it is better to say so than let a detector guess.

Our advice: declare every release yourself, through your distributor, using YouTube's own examples, and keep a dated note of what AI did. If YouTube's label is ever tied to reach or royalties, as TIDAL has already done for music it identifies as wholly AI-generated, that note will matter.