What our review checks
The automated checks every release runs at submit, the difference between a hard failure and a flag, and why a person approves every release before it ships.
Every release is checked before anything leaves Run. The checks start the moment you submit and run in three legs: your metadata, your audio and artwork files, and a content match against recordings that are already released.
Nothing reaches a store on automation alone. A release that clears the automated checks does not ship straight out: it waits for a person on our team to approve it, with anything the checks flagged attached for them to read.
Two words are worth learning, because they decide what happens to your release. A hard failure stops it and sends it back to you with the reason written out. A flag stops nothing: it rides along to the person reviewing the release, who decides.
The metadata check
This one reads every piece of text that travels with your music: titles, artist credits, genre, language, track order, dates, the barcode and ISRCs, the people you credited, the publishing splits and the AI disclosures.
It looks for four kinds of problem:
- Wording stores refuse. Search bait in a title (“type beat”, “official audio”), karaoke and “originally performed by” versions, “in the style of” phrasing, and “AI cover” or “AI voice” versions of somebody else’s work.
- Credit lines that will land wrong. At least one main artist, at most 4, nobody credited twice, and any artist who told us they already have store pages must have one connected. Without that, a release lands on a stranger’s page or mints a duplicate one.
- Codes. The barcode and the ISRCs have to be the right shape, and neither may already belong to another release or recording in our catalogue.
- Completeness and consistency. Every track needs a performer, a composition and lyrics credit, and a production and engineering credit; at least one credited person must not be fully AI; publishing splits, if you gave any, must total exactly 100. On top of that it looks for contradictions: an explicit flag that disagrees with the title, a credit marked as AI-involved while both disclosures say none, two artist profiles on your account with almost the same name, a title you have already released.
The full rules are in metadata rules and credits and splits.
The audio and artwork check
Each master is decoded and measured. Some of it is a hard gate, taken from what the stores themselves accept: a lossless file (WAV at 16, 24 or 32-bit, or 16-bit FLAC), 44.1 kHz to 192 kHz, stereo, at least one second long, no larger than 2 GB. A file that misses one of those comes straight back.
The rest is measured and flagged rather than blocked: true peak above -1 dBTP, loudness far above the level stores normalise to (around -14 LUFS), more than two seconds of silence at the start or end, two identical channels (a mono recording inside a stereo file), and a spectrum that looks like it was made from an MP3 or AAC. That last one is a net, not a proof, which is exactly why it flags instead of failing.
Cover art is a hard gate: square, at least 3000 by 3000 pixels, no transparency, and readable as an image. See audio requirements and artwork requirements.
The content match
Every track is fingerprinted from a short excerpt and compared against a large catalogue of released recordings. Two kinds of match come back, and they mean different things:
- The same recording. Your file matches a master that is already out. That needs explaining, and a reviewer looks at it.
- The same composition, performed differently. This is what a legitimate cover looks like from the outside, and it is the expected result when you have declared one.
A match never rejects a release on its own. It is a flag, and a person decides between “a cover with the paperwork attached” and “somebody else’s recording”. If the scan cannot run at all, we record that it did not run rather than treating the track as cleared, so the human rights check still happens.
What you declare on the track is what makes this quick. See covers, remixes and samples.
What each outcome does
| Outcome | Your release reads | What happens next |
|---|---|---|
| At least one hard failure | Checks failed | It comes back to you with every reason listed. Reopen, fix, submit again |
| Flags only | In review | It goes to a person with the flags attached |
| Nothing found | In review | It still goes to a person: approval is what starts delivery |
| Approved | Approved | Delivery begins, and you can follow it store by store |
We work the review queue oldest first. You get an email each time the state changes, so there is nothing to sit and watch. The findings themselves, and what each one wants from you, are listed in common rejection reasons.
What the checks do not do
They do not judge your music, your genre or your artwork taste. They cannot see rights you have not declared, which is why the track questions about covers, public domain and samples matter more than they look. And they are not the stores’ own review: a store can still come back with feedback after we have delivered, which shows up on the release. See tracking delivery.