How to release music on Spotify, and the AI rules that changed

You cannot upload directly to Spotify. That surprises people every time, and it is the first thing to understand: streaming services take music from distributors, not from artists. Your job is to pick a distributor, deliver correctly formatted files and metadata, and wait.

Here is the actual sequence, the deadlines that matter, and the AI-specific rules that have tightened significantly.

The sequence, in order

1. Finish and master the track. Deliver the best audio you can. Most distributors want WAV at 16-bit 44.1 kHz or better. Mastering is the stage that gets it there, and free automated mastering is genuinely adequate for a first release.

2. Prepare artwork. 3000 by 3000 pixels, square, RGB, no web addresses, no logos, no third-party imagery you do not own. The full rejection list is worth checking before submission rather than after.

3. Pick a distributor. The main models: annual subscription for unlimited releases, per-release fees, or free distribution with a revenue share. Which is cheapest depends entirely on how much you release. One track a year favours pay-per-release; twelve favours a subscription.

4. Register your artist name and check it is free. Availability across streaming services is a discovery question, not a legal one, and it is easier to fix before release than after.

5. Prepare metadata. Artist name spelled identically everywhere, track title, genre, language, contributor credits, and an ISRC code, which most distributors generate for you.

6. Set a release date at least three to four weeks out. This is the deadline people miss. Editorial playlist pitching typically requires submission well in advance, and a track submitted for release tomorrow cannot be pitched at all.

7. Submit and wait. Delivery to stores takes days, and rejection for artwork or metadata problems costs you the whole window.

What actually gets releases rejected

In rough order of frequency:

Artwork violations. Text that does not match the metadata, promotional URLs, imagery you do not own, wrong colour space.

Metadata inconsistency. Artist name spelled differently between the submission and the artwork, or between releases, which fragments your artist profile.

Rights problems. Covers without a mechanical licence, samples without clearance, and anything containing material you cannot document ownership of. The two copyrights in every song explain why a cover needs permission even though the recording is yours.

Misleading titles. Naming another artist in your title or metadata to catch their traffic gets caught and penalised.

The AI rules, which have changed a lot

This is where recent policy has moved fastest, and where advice written a year ago is now wrong.

Platforms are detecting and tagging. Deezer reported receiving close to 75,000 fully AI-generated tracks a day by mid-2026 and tags them, excluding tagged tracks from algorithmic recommendations and editorial playlists. It also reported that a large majority of streams on fully AI-generated songs were identified as fraudulent and demonetised. How detection works and how each platform differs is worth reading before you plan a release schedule.

Other platforms differ. Some lean toward labelling rather than exclusion, at least one bans AI-generated music outright, and several encourage or require disclosure.

Distributors ask. Several now require you to declare AI involvement at submission, and misdeclaring is a terms violation.

Volume uploading is actively policed. Mass-delivered generated catalogues are the specific behaviour these systems were built to catch, and the economics of that model have collapsed accordingly.

Your rights must be clean before you submit. On Suno, commercial rights attach only to songs made while you held an active paid plan, never retroactively. Distributing a free-tier track is a licence violation regardless of whether anyone notices. The rule in full takes five minutes.

The honest summary: releasing generated music is legitimate if the rights are clean and you disclose. Releasing a hundred tracks a month to farm streams is a business model with an enforcement mechanism pointed at it.

What to do on release day and after

Do not expect the algorithm. For a first release without an audience, algorithmic promotion is not coming, and it certainly is not coming for a tagged AI track on a platform that excludes them.

Get the link out where you already have people. The playlist pitch matters far less than the twenty people who will actually listen because they know you.

Claim your artist profile on the major services. This is a separate step from distribution and it is what lets you control your bio, image and links.

Release consistently rather than perfectly. A recognisable sound across a series does more than one polished single, which is why building a house sound and reusing style recipes matters more than agonising over one track. Two or three recipes reused across a year is the whole technique, and our catalogue has 1,960 of them with ten open free.

A realistic first-release timeline

• Week 1: finish the track, master it, make artwork
• Week 2: choose a distributor, sort metadata, check the artist name is free
• Week 3: submit with a release date four weeks out
• Weeks 4 to 6: pitch for editorial consideration if your distributor supports it, prepare what you will post
• Release day: send the link to people directly
• After: start the next track, because the second release is what makes the first one findable

FAQ

Can I upload music to Spotify directly? No. Spotify takes music from distributors. You need one.

How much does it cost to release a song? Depends on the model: an annual subscription, a per-release fee, or free with a revenue share. Release frequency decides which is cheapest.

How long does it take? Plan three to four weeks ahead. Delivery itself takes days, but editorial pitching requires lead time.

Can I release AI-generated music on Spotify? With clean rights, yes, though platforms are detecting and tagging generated music and several exclude tagged tracks from recommendations. Disclose when asked.

What gets a release rejected? Artwork violations, inconsistent metadata, and rights problems with covers or samples.

Do I need to master before releasing? Yes, at least lightly. Automated mastering is adequate for a first release.