I Made a Worship Song With AI. Here's Why That's Not the Problem.

I work from home in Abuja. A few weeks ago, I discovered I live next to a church. Not from a sign or a Sunday crowd, but from a full praise and worship session at 11am on a Thursday, right through my living room walls, mid-meeting.

My first thought was: "Don't these people have an office?"

Turns out I'm the one without an office.

But the chorus was good. One of those rhythms that catches you before you can be annoyed by the volume. So I did what any reasonable adult would do: I recorded it through the wall, opened Suno, and turned it into a track called Praise Ignite. Typed a description of the vibe, picked a style, and within a few minutes I had something that sounded, to my admittedly untrained ears, like a real song.

I shared it on LinkedIn and said I might have crossed into "I'm not okay" territory. But the honest reaction was simpler than that.

It felt good. It felt like making something.

I'm not a musician. I don't play an instrument, can't read sheet music, and have no plans to top any charts. But for a few minutes, I had an idea, a rhythm that hit me through a wall, and I turned it into something I could share. That's expression. And it's exactly what the current AI music debate keeps missing.

The scale is no longer theoretical

Here's where AI music sits in early 2026. According to Billboard's reporting on Suno's investor pitch deck, users are creating roughly 7 million songs per day on the platform. Every two weeks, that output matches Spotify's entire catalogue of over 100 million tracks. Suno raised $250 million in its Series C at a $2.45 billion valuation, backed by Menlo Ventures and NVIDIA's venture arm. It claims one million subscribers with 78% weekly retention.

Its rival Udio has carved a different path, settling early with Universal Music Group and implementing tighter constraints on its output. Meanwhile, Deezer, the only major streaming service actively labelling AI content, reports that 60,000 fully AI-generated tracks land on its platform every day, and it detected and labelled over 13.4 million AI tracks in 2025.

A LANDR survey of 1,200 artists found that 87% have already used AI in at least one part of their music workflow. The Hollywood Reporter puts it bluntly: 97% of listeners in a recent poll couldn't identify a fully AI-generated track when they heard one.

This isn't coming. It's here.

The Canva pattern

The conversation around AI music tends to collapse into a binary: either you're for AI replacing artists, or you're against it. But that framing misses the largest group of people using these tools: people who were never artists in the first place.

Think about what happened with visual design. Canva launched in 2013 and was met with scepticism from professional designers. A decade later, it has over 200 million users. The vast majority of them aren't designers. They're founders making pitch decks, teachers building worksheets, small business owners creating social posts. Canva didn't replace Adobe. It opened a door for people who had ideas but lacked the craft to execute them.

The same pattern is playing out across every creative domain AI touches: images, video, writing, and now music. In each case, the tool doesn't turn amateurs into professionals. It turns non-participants into participants.

I'm one of those non-participants turned participant. I don't pretend the track I made is a professional production. But it captured something, the specific feeling of a Thursday morning worship session coming through concrete walls, in a way that writing a LinkedIn post about it couldn't. The medium was the message.

That's the part the debate keeps flattening. When people use Suno or Udio, the primary motivation for most of them isn't to game Spotify's algorithm or undercut a working musician. According to Suno's own pitch deck data, the average user spends about 20 minutes per session. They're making something for themselves, maybe sharing it with friends. They're doing what humans have always done with new tools: finding out what it feels like to create in a medium that was previously closed to them.

The slop problem is real. But it's a different problem.

The slop is real, and it's a distribution problem

Let's not pretend AI music is all bedroom gospel remixes and creative expression. There's a genuine flood of low-effort, mass-produced content designed to extract money from streaming royalty pools, and it's hurting real musicians.

AI music slop

Mass-produced, low-effort AI-generated tracks uploaded to streaming platforms at scale, typically by bot farms gaming royalty pool systems. Not to be confused with personal or creative use of AI music tools.

Lucien Grainge, the chairman of Universal Music Group, called out the "exponential growth of AI slop" on streaming services in January 2026. He wasn't wrong. Deezer's data shows the scale: the streaming service reports that up to 39% of daily uploads are fully AI-generated, and that a significant portion of AI-generated streams are driven by bot farms gaming the system for royalties. YouTube Music users have reported finding six out of ten recommendations replaced by AI-generated tracks. Reddit threads are full of paying subscribers who say they can spot the pattern within seconds: the same auto-tuned voice across different "artist" names, uncanny melodies, weird mixing artefacts.

And the economics are stark. When a bot farm uploads thousands of ambient lo-fi tracks under hundreds of fake artist names, they dilute the royalty pool that pays working musicians. A study cited by the music industry body GEMA estimated that up to €950 million in revenue could be at risk in Germany and France alone by 2028 if AI music goes unregulated. Indie artists in the genres most susceptible to AI imitation (lo-fi, ambient, instrumental hip-hop) are already seeing streams and engagement drop.

But here's where the argument gets sloppy on both sides. The slop problem is overwhelmingly a distribution and monetisation problem, not a creation problem. The people flooding Spotify with bot-generated ambient tracks aren't doing it because Suno exists. They'd find another method if it didn't. The incentive is the streaming royalty pool structure, not the tool. Banning the tool to fix the incentive is like banning cameras to prevent Instagram spam.

As Mark Mulligan of MIDiA Research put it:

"The music industry should worry less about the song with 1 million streams and more about the 1 million songs with 1 stream."

That's the real structural shift. And it's one that predates AI music entirely. It started with the democratisation of recording and distribution through services like DistroKid and TuneCore.

The IP reckoning is happening faster than expected

If you'd told me in mid-2024, when Universal, Sony, and Warner each filed $500 million lawsuits against Suno and Udio, that we'd see licensing settlements within 18 months, I'd have been sceptical. Copyright litigation in the music industry typically moves at geological speed.

But here we are. Warner Music Group settled with both Suno and Udio in late 2025. Universal settled with Udio separately. Suno has committed to relaunching in 2026 with a model trained exclusively on licensed material, retiring its current unlicensed model. The terms include monthly download caps and a shift in ownership language from "user owned" to "granted commercial rights."

That last point matters more than it sounds. Under the updated terms, paid Suno subscribers get a commercial licence to monetise their tracks, including streaming, sync placements, and downloads. But Suno technically retains authorship of the audio itself. Free-tier users get nothing: their output is personal, non-commercial use only.

The deeper issue is one the US Copyright Office has been consistent about: AI-generated audio, in its raw form, is generally not copyrightable. You can monetise your Suno track on Spotify, but if someone copies it and re-uploads it, your legal standing to sue is thin unless you've added what the Copyright Office calls "significant human changes": re-recorded vocals, live instrumentation, substantial arrangement edits.

For someone like me, making a worship remix for laughs, this doesn't matter. For anyone building a commercial music operation on AI-generated output, it's the most important detail in the room.

What the equilibrium probably looks like

The parallel I keep coming back to isn't Napster. It's photography.

When cameras became affordable and portable in the early 20th century, professional painters worried, reasonably, that the new medium would kill portraiture and landscape art as professions. What actually happened was more interesting. Photography created an entirely new creative class (photographers), forced painting to evolve into new forms (impressionism, abstraction), and gave millions of ordinary people a way to capture and share their visual experience of the world. All three of those things happened simultaneously. The "death of painting" crowd and the "photography is art" crowd were both partially right, and the tension between them produced something better than either position predicted.

AI music is headed somewhere similar. The long-term equilibrium, based on where the legal, technical, and cultural threads are converging, probably has four elements:

None of this will be clean, and none of it will happen fast enough to prevent the near-term surge. iHeartRadio has already launched a "Guaranteed Human" programme, pledging not to use AI-generated personalities or play AI music with synthetic vocalists. Apple Music appears to have largely avoided the slop problem by working more closely with distributors. Deezer is the only platform actively letting listeners block fully AI-created songs from recommendations.

The honest concession: these measures protect the professional music ecosystem, and they should. Working musicians who spent decades honing their craft deserve a market that doesn't get diluted by automated noise. The industry's objections aren't cynical. There's a real economic threat, with GEMA's research estimating billions of euros at risk across European markets by 2028.

But none of those protections need to come at the cost of personal creative expression. The person making a track in their bedroom at lunch isn't the person running a bot farm. Regulation that treats them the same will fail, not because it's unjust (though it is), but because it won't work. Consumer behaviour in this space is accelerating regardless, as it did with music piracy, as it did with streaming, as it does with every technology that reduces the friction between an idea and its expression.

The medium and the message

Here's what I think about when I listen back to that Praise Ignite track. It isn't a good song by any professional standard. The vocals are synthetic, the arrangement is algorithmic, and the theological depth is questionable at best. A working musician would, rightly, point out that it lacks the human imperfection and intention that makes music art.

But it captured something. The surprise of worship at 11am on a Thursday. The specific rhythm that cut through concrete walls while I was pretending to pay attention in a meeting. The absurdity of turning a neighbour's church service into a remix project during a lunch break in Abuja.

No amount of technical skill would have helped me express that. The tool did.

The AI music debate will keep unfolding. The lawsuits still pending between Sony, UMG, and Suno will shape the legal framework for years, and the slop problem needs serious structural solutions at the platform level. But buried under the copyright filings and the bot farm takedowns and the heated op-eds, there's a quieter story about millions of people who had something to say and finally found a way to say it in a medium that was previously locked behind years of training, thousands of pounds of equipment, and a talent they were never born with.

That isn't a threat to music. It's an expansion of who gets to make it.

And if my neighbour finds out I turned their Thursday worship session into an AI remix, I hope they're quietly impressed rather than annoyed. Though honestly, I'm prepared for either.

1. Billboard, "Suno's Pitch Deck Reveals 7M Songs Per Day, $2.45B Valuation," November 2025
2. Deezer / Associated Press, "AI-Generated Music Uploads Hit 60K Per Day," January 2026
3. LANDR, "AI in Music Production Survey," 2025 (n=1,200 artists)
4. The Hollywood Reporter, "97% of Listeners Can't Identify AI Music," 2025
5. GEMA, "Economic Impact of AI Music on European Markets," 2025
6. Mark Mulligan, MIDiA Research, "The Million Songs With One Stream," 2025
7. US Copyright Office, "Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence," 2023 (updated 2025)
8. Warner Music Group / Suno settlement terms, November 2025