How to create and sell AI-generated music online

Making a track is the easy part now. Turning a folder of tracks into income is the part nobody explains. This is that part.

Published 2026-07-30 by Cat Claw, the UK's create-and-distribute AI studio. About a 14 minute read.

The short version

• AI music is legal to sell in most places, but ownership rules differ by country and the platform terms decide what you actually hold.
• Niche beats volume. A creator known for cinematic horror outsells one with 200 unrelated tracks.
• Royalty-free standard licences run roughly $5 to $30. Exclusive rights run into the hundreds. Sync pays most.
• Bundle related tracks into themed packs. Buyers pay more for a coherent set than for a better single track.
• Streaming is a visibility channel, not an income one. Treat licensing as the business.

AI music has stopped being a novelty and started being a supply chain. Video editors need background beds. Podcasters need intros. Brands need audio identity. All of them are buying, and most of them are buying from independents rather than libraries.

You can make a usable track in minutes without reading a bar of sheet music. What decides whether you earn anything is everything that happens after: what you make, who it's for, how you licence it and where you list it. This guide covers the whole path, from first track to first sale.

What is AI-generated music and why does it matter?

Quick answer. AI-generated music is audio composed by a machine-learning model rather than played or written by hand. The model learns patterns in melody, rhythm, harmony and structure from large collections of music, then builds a new original track from your prompt or settings. The output isn't a copy of any existing song, which is what makes it sellable.

Quality has moved fast enough that in normal listening conditions most people can't pick an AI-composed track out of a playlist. That's the shift that turned it from a toy into a market.

• $3bn+ Value of the global AI music market in 2025 (Grand View Research)
• 28%+ Projected annual growth rate through to 2030 (Grand View Research)

For content creators it matters because licensed music is expensive and slow to source. For brands it means custom audio without commissioning a composer. For independents it's an income stream with almost no upfront cost, which is rare.

Do you need musical experience?

No. You don't need music theory, sheet music or a digital audio workstation. You describe what you want in plain language and you get a track back. That's the whole barrier gone.

For experienced producers the tools work differently: less a replacement, more a fast way to generate raw material and get past a blank session. Either way, the skill that separates sellers from hobbyists isn't production, it's taste. Knowing which of 10 tracks is the one worth listing is the job.

What types of music can AI generate?

Effectively every genre: ambient, electronic, hip-hop, classical, jazz, pop, lo-fi, cinematic and everything between. You can steer mood, energy, instrumentation and length, and you can generate with vocals or fully instrumental.

The commercially useful work clusters tightly, though. Background beds for YouTube, podcast intros and outros, loops for Reels and Shorts, corporate presentation music, game soundtracks and audio branding. Full albums get made, but functional tracks are what sell repeatedly.

How does AI music creation actually work?

You give the model an input, it generates audio that fits. Most platforms take one of 2 approaches, and the better ones take both.

• Text-to-music: you type a description like "upbeat jazz with a light piano melody, warm and unhurried" and get a track built to match.
• Style-based: you pick a genre, mood or reference and the model works inside that frame.

In practice the controls that matter are a style prompt, your own lyrics if you want vocals, and an instrumental toggle for when you don't. That's the setup in Cat Claw's Music Studio: describe the sound, paste lyrics or leave them out, and get a full song back, up to 5 minutes. Finished tracks land in a gallery you can play from and download later.

Generation takes seconds to a couple of minutes depending on length. Then you download, regenerate variations, or change one input and go again. That last loop is where the actual craft lives.

How to make a track worth selling

Most people who fail at this generate 50 tracks in an afternoon, list all 50, and sell none. The tracks that sell are made deliberately, for somebody.

1. Start with the buyer, not the sound. "Lo-fi track" is a hobby. "90-second lo-fi bed for a productivity YouTuber, no vocals, nothing that fights a voiceover" is a product. Write the brief before you write the prompt, and the prompt mostly writes itself.
2. Be specific in the style prompt. Name the genre, the tempo feel, the key instruments, the mood and the use. "Warm analogue lo-fi, 75bpm, soft Rhodes and brushed drums, mellow and unobtrusive, for study videos" gets you far closer than "chill beats".
3. Decide vocals early. Vocals make a track memorable and much harder to licence, because they compete with whatever the buyer puts on top. For functional catalogue work, instrumental sells more. Keep the vocal versions for release, not licensing.
4. Generate several, then be ruthless. Run the same brief a few times with one variable changed. Small shifts in tempo, key or instrumentation give dramatically different results. Keep notes on what worked: that's how you build a repeatable sound rather than getting lucky.
5. Reject most of them. Listen critically and bin anything generic or technically off. Buyers can tell the difference between a curated catalogue and a bulk upload, and one weak track on your page devalues the strong ones next to it.

A useful test before listing anything: play it under a voiceover or a rough video cut. If it fights the thing it's meant to support, it's a nice track and a bad product. Functional music has a job, and the job is to not be noticed.

Where can you sell AI-generated music?

Where you list depends on what you make and how you want to get paid. These are the routes worth knowing.

One thing to be clear about. Cat Claw is where you make the music, not where you sell it. The Cat Claw Prompt Marketplace trades image and video prompts, not audio tracks, so plan to list finished music on the stock, direct or sync routes above. Read any platform's AI submission policy before you upload, because they differ and they change.

How do royalties and licensing work?

The licence is the product. Same audio file, 3 different prices depending on what the buyer is allowed to do with it.

Before listing anything, be confident you have the right to sell it. Most AI music tools grant commercial rights to the person who generated the track, but the detail varies. Read the terms of whatever you used, specifically around ownership, distribution and whether the platform keeps any claim on the output.

Can you monetise AI music on streaming platforms?

Yes, with conditions. Spotify, Apple Music and Amazon Music all accept AI-generated music provided it meets their general content standards, and you'll need a distributor such as DistroKid, TuneCore or CD Baby to get it there.

Disclosure requirements are tightening. Spotify has said tracks must be labelled as AI-generated where they imitate a specific artist in a way that could mislead listeners, and distributors are adding their own rules. This area is moving quickly, so check your distributor's current guidance rather than relying on what was true last year.

Set expectations on the money. Per-stream payouts are tiny, and a catalogue of functional instrumentals is not what streaming rewards. Use it for discovery and credibility, and earn from licensing.

How to price your AI music

Price too low and you signal low quality while leaving money behind. Price too high with no track record and nothing moves. Start by looking at what comparable tracks go for on the platform you're listing on, then adjust for what you're actually offering.

• Length and completeness: a 60-second social loop is worth less than a 3-minute composition delivered with stems.
• Exclusivity: an exclusive licence should cost multiples of a standard one, because you're giving up every future sale of that track.
• Usage breadth: broadcast and paid-advertising rights price higher than a personal-project licence.
• Bundling: themed packs of 5 or 10 related tracks raise both perceived value and average order size. Buyers building consistent audio branding actively prefer them, because it saves searching.

As your catalogue and reputation grow, raise prices on new releases and leave the older ones at entry rates. The cheap back catalogue becomes the thing that brings new buyers in.

The economics of building a catalogue

This is the part most guides skip, and it's the part that decides whether this is a business or a hobby. You need to know your cost per track and how many you have to sell to clear it.

Generation cost is the easy half. In Cat Claw's Music Studio a song costs 8 credits, whichever engine you use, so the entry plan at £14.99 a month covers roughly 43 songs. If you reject 3 out of every 4, which is a healthy ratio, that's still about 10 listable tracks a month for under £15.

Run the numbers before you commit. 10 listable tracks a month at a $5 to $30 royalty-free licence means a single sale roughly covers the month's generation. Everything after that is margin. The real cost isn't credits, it's the hours you spend curating and marketing, so price your time in when you decide whether a niche is worth entering.

Two things protect that maths. Generate in batches against a single brief so you're comparing like with like, and keep every usable track even if you don't list it immediately. A B-side that doesn't suit today's pack often anchors next quarter's.

Marketing your music to the right audience

Making good tracks is half the work. The other half is being findable by the people who need them, and they're easier to locate than most sellers assume.

Video creators on YouTube, TikTok and Instagram are the most active buyers of background music and sound design. So are podcasters and streamers. They gather in creator communities on Reddit, Discord and Facebook, and sharing short clips of your tracks there, with a clear line about where to licence them, drives steadier traffic than any ad.

Showing your process works particularly well for AI music. A short video of a track being built attracts buyers and other creators at the same time, and it positions you as someone who knows the craft rather than someone dumping output. Posting those clips across several platforms at once, rather than building a separate plan per channel, is the difference between a habit you keep and one you abandon by week 3.

Consistency beats volume. Turning up in the same places every week builds name recognition, and a catalogue that keeps growing stays visible in platform search.

How can AI music creators build a following?

Start with a recognisable identity. Pick a lane, whether that's cinematic ambient, workout energy or lo-fi study beats, and produce inside it consistently. Buyers follow creators whose next release they can predict, because predictability is exactly what a repeat buyer wants.

Engage rather than broadcast. Answer questions, give something useful before asking for a sale, and be a member of the community you're selling into. Then build an email list, because it's the one audience no platform can take away from you. Use it for new releases, early access to packs, and the occasional look behind the process.

Is selling AI-generated music legal?

In most places, yes. The law is still settling, and it differs by country, so it's worth understanding the shape of it before you list.

In the UK, copyright normally requires human authorship. But the Copyright, Designs and Patents Act 1988 includes a provision for computer-generated works, granting copyright to the person who made the arrangements for the work to be created. If you're the one prompting and directing the tool, that provision may cover you. The UK Intellectual Property Office has published guidance on AI and IP, and it's worth reading directly.

In the US, the Copyright Office has said purely AI-generated work with no human creative input isn't eligible for copyright, while work with sufficient human creative contribution may be. That line is being actively argued.

• Comply with the terms of whatever tool generated the track, particularly on ownership and resale.
• Be straight with buyers about how the music was made. Some licensing platforms require it, and it avoids a dispute later.
• Don't market a track as sounding like a named artist. That's where imitation claims and platform disclosure rules bite hardest.
• If you're building a serious commercial operation on this, take advice from an IP solicitor. This guide isn't legal advice.

How to stand out as the market fills up

Competition is rising quickly, and generating tracks is no longer a differentiator because everyone can do it. These are what still work.

• Specialise hard. A creator known for high-quality cinematic horror scoring attracts a more loyal and better-paying audience than a generalist, and niche terms are far easier to rank for in platform search.
• Curate visibly. A small catalogue where everything is good outperforms a big one where most things are fine.
• Write proper metadata. Accurate tags, a descriptive title and the right category are how buyers find you at all. This is unglamorous and it's most of the job.
• Keep prompt notes. The combinations that produce your best results are your actual intellectual property, more than any single track.
• Collect feedback and use it. Even short testimonials tell a new buyer that someone else took the risk first.

Frequently asked questions

What is AI-generated music and how is it made?

Audio created by machine-learning models trained on large collections of existing music. You give a prompt or set parameters such as genre, mood and tempo, and the model generates a new original track. No instruments, no recording equipment and no music theory required.

Do I own the copyright to music I create with AI tools?

It depends on the platform and your country. In the UK, computer-generated works can receive a form of protection under the Copyright, Designs and Patents Act 1988, which grants copyright to the person who made the arrangements for the work's creation. In the US, purely AI-generated content is generally not eligible. Always read the terms of the specific tool you used before selling.

Can I sell AI music with no production experience?

Yes. Modern tools need no music theory or DAW knowledge. If you can describe a mood and a genre in plain language you can make a sellable track. The skill that matters commercially is curation: knowing which of your tracks is actually worth listing.

Where is the best place to sell AI-generated tracks?

Stock music marketplaces like Pond5 and AudioJungle are the usual starting point because the buyers are already there. Direct licensing from your own site pays the best margins once you have traffic. Sync and brand work pays most per track but takes relationships and a credible catalogue first.

Can I sell music on the Cat Claw marketplace?

Not currently. The Cat Claw Prompt Marketplace trades image and video prompts rather than audio tracks. Cat Claw's Music Studio is where you make the music; list the finished tracks on stock music sites, your own storefront, or through sync and distribution partners.

How much money can you realistically make selling AI music?

It varies enormously and depends far more on marketing than on output. Creators with a focused catalogue and consistent promotion report anywhere from a few hundred to several thousand pounds a month, and the ones who do best combine stock licensing, sync work and streaming rather than relying on one channel. Treat early months as building a catalogue, not earning from one.

How much does it cost to generate a track?

In Cat Claw's Music Studio a song costs 8 credits on either engine, so the £14.99 entry plan covers roughly 43 songs a month. Assuming you reject most of what you generate, which you should, that's around 10 listable tracks a month for under £15 in generation cost.

Should my tracks have vocals or be instrumental?

For licensing, instrumental sells more. Vocals compete with whatever the buyer lays over the top, whether that's a voiceover or dialogue, so functional catalogue work is usually cleaner without them. Keep vocal tracks for releases where the song itself is the product.

Is AI-generated music allowed on Spotify and other streaming services?

Yes, with conditions. Major platforms accept AI-generated music distributed through an approved distributor. Some require disclosure where a track imitates a specific artist. The rules are changing quickly, so check your distributor's and the platform's current guidance before submitting.

How long should a track be?

Match it to the use. Social loops run 15 to 60 seconds, YouTube background beds 2 to 3 minutes, and podcast intros often under 30. Producing a long version plus edited cutdowns from the same idea gives you several products from one good track.

What makes one AI music track sell better than another?

Filling a specific, underserved need. Good audio quality, accurate metadata, a clear title and the right niche category all help buyers find it. Bundling into themed packs and releasing on a consistent schedule lifts sales over time more than any individual track does.

Do I have to tell buyers the music is AI-generated?

On many platforms yes, and it's sensible everywhere. Some licensing sites require it in submission terms and some streaming platforms require labelling in specific cases. Beyond the rules, being upfront avoids disputes after a sale, and buyers who care will ask eventually.

Make the track. Then go and sell it.

Cat Claw's Music Studio turns a style prompt and your lyrics into a finished song, up to 5 minutes, vocals or instrumental. Free to sign up, no card needed.

• Start using it now
• More Cat Claw guides
• Prompt Marketplace · Legal & policies

Make tracks worth paying for, work out what to charge, and get them licensed. A practical guide to building an AI music catalogue that actually earns.