Apple Music Never Banned AI. It Started Labelling It.

The confusion in this topic comes from treating one story as two, or two stories as one. They are genuinely separate events, six months apart, and they do different jobs.

On 5 March 2026 Apple told partners it was adding Transparency Tags to its music delivery system, letting labels and distributors declare where AI had been used. On 20 August 2026 Apple emailed partners again to say that the information collected by those tags would become visible to listeners as Made With AI labels later in the year. The second announcement added no new obligation. It changed who gets to see what the first one collected.

Read down that column and the shape of the sector in 2026 is obvious. Every major store has decided the answer to AI music is metadata rather than a ban — and each one is building its own.

Apple's trigger is that a tag should be applied when a material portion of the content was created using AI. There is no published percentage, no stem count, no duration threshold. People keep looking for the number. There isn't one.

That reads like an oversight until you notice how Apple framed it: similar to genres, credits, and other metadata, Apple defers to content providers to determine what qualifies as AI content. Genre is the giveaway. Nobody has ever published a percentage for when a track becomes house rather than techno; the person delivering it decides, and the store takes them at their word. Apple has put AI disclosure in that same category deliberately.

Which leads to the part worth being blunt about: nothing in this system inspects your audio. Apple is not running a classifier over deliveries to catch undeclared AI. Trade coverage at the time of the March announcement noted that how Apple would detect undeclared AI use was unclear, and nothing published since has clarified it. Every label on the platform is only as accurate as the person who submitted it.

That right-hand column is not Apple's published list — Apple has not published one — but it follows the logic of the wording, and it matches how YouTube draws the same line, where “using effects to enhance previously recorded audio” is explicitly outside the disclosure requirement. Two platforms, two schemes, the same instinct: generation is disclosable, processing is not.

Here is the structural fact that makes this whole topic easier to act on. Apple's Transparency Tags are part of the delivery specification — the package a label or distributor sends to Apple. An independent artist has no interface to Apple Music in which to set them. Neither does Spotify's DDEX credit reach the store any other way.

So both disclosures come down to fields in your distributor's upload form, and whether that distributor supports them yet. If those fields are absent, or you leave them blank, your release goes out undeclared — not through any decision of yours, but because nobody asked. This is the failure mode worth checking for before your next delivery rather than after it.

It is also why distributor choice quietly became the most consequential decision in AI music releasing. DistroKid was the first distribution partner for Spotify's AI credits in April 2026 and publishes explicit written rules for AI content; we go through what its screening accepts and rejects in the DistroKid and AI music guide. The mechanics of getting a generated track through that pipeline are in our Suno to Spotify walkthrough.

Step three is the one creators most often collapse into step two, so it is worth stating plainly: cleaning a file has no effect on whether it must be disclosed. Disclosure attaches to how the music was made. Cleanup attaches to how the file sounds and whether it passes a distributor's technical screening. They are different problems that happen to land in the same week, and the guide to AI audio cleanup covers the second one properly.