Wanna plai?

Every platform treats AI differently.

Sep 22, 2026. Disclosure. Spam farms. Persona packaging. Human still needs a lane. Figures below are reported, not PLAIGROUND lab results.

Listeners follow a face

Streaming still runs on people. Not files.

A WAV does not get saved. A person does. The algorithm rewards saves, repeats, playlist adds, “who is this?” searches. That is why a convincing identity can make a generated track look like an act. That is why a raw dump dies in the long tail. The hard part: every store treats AI differently, and the rules move. Artists need a distributor that handles disclosure and does not wink.

Deezer / Qobuz / Tidal

Reported as the hard line. Deezer scans for artifacts, tags AI, kicks tracks out of recs, and has demonetized a large share of those streams. They have claimed high accuracy on fully synthetic audio. Qobuz has moved the same way. Tidal has signaled it will stop paying fully AI and auto-tag it. Listing is not a paycheck.

Spotify

A different route. Reported: tens of millions of spammy tracks removed. Filters on upload patterns and duplicates. Labeling, including AI Persona–style framing, that can keep catalogs out of personalized recs. They still lean on voluntary disclosure from distributors. That only works if the distributor enforces it.

The arms race

Detection mixes spectral analysis, timing, metadata, watermarks. Tools exist to strip artifacts. Compression or a pitch shift fools a lot of detectors. The bottleneck is not magic tech. It is enforcement at the distributor. Most claim to block AI spam. Many still let it through.

Slop is most of the volume. Not most of the people.

By artist behavior, reported research says only a few percent of AI-associated acts flood at farm volume, and those accounts can produce the majority of AI track volume. The other ~97% are lower-volume creators.
By track outcomes, figures often cited put ~93% of AI tracks under a thousand plays. Platforms have also claimed a large share of AI streams look fraudulent.
Among AI music that actually gets released and charted, only a minority is described as fully machine-generated. The rest has a human in the loop.
Kick spam farms. Do not punish every indie who used a tool.

Why a wrapper works when it works

Cases people point to, Xania Monet, Breaking Rust / Velvet Sundown, Enlly Blue, did not win because a file was clever. They won because there was a wrapper. Consistent face. Bio. Visual language. A catalog that felt like one act. Platform-native clips first. Then the DSP. Without that wrapper, most AI tracks sit in the 93%. With it, a few break into the sliver that earns.

A human core. Lyrics or a story from someone’s life. Not “write a sad R&B song.”
One visual language. Same face, lighting, era across art and Reels.
An artist page that behaves like an artist page. First-person bio. Studio-style clips. Not “AI music by username.”
Platform-native first. The hook on TikTok or Reels. Then the full song in stores.
Consistency over volume. A few tracks that belong together. Not 30 genres in a month.

What we do. What we are building.

We are not promising we can detect every AI stem better than Deezer. Soft-stop that hype.
Live now: mandatory class at upload. No AI. AI-assisted. Fully AI. Human stays Human. We will not release a human record as AI.
Live now: no clones. No dead-singer covers. Split sheet before delivery. 0% royalty cut.
Building toward: tighter farm caps, per-store routing with a warning before a bounce, a persona kit that is not a raw dump into the main catalog.
We do not just distribute files. We help distribute acts, with honest labels.

Every platform treats AI differently. You attest the label. We pass it on. Stores decide pay and playlist rules. Human artists stay Human. Their own lane. Honest labels. No bury-under-AI.

Truth for artists · Human ships as human · Nobody marketed it · Release Now

Tell the truth. Then push the button.

Upload once. Get paid everywhere. PLAIGROUND takes none of the royalties.

Release Now