AI music: what it is, what it isn’t, and where I stand

AI music is everywhere right now, and a lot of people don’t trust it. I use AI in my music, so I think it’s only right that I’m upfront about how I use it and how I feel about it.

This first article covers three things: what people think and feel about AI music, what the music industry is trying to do about it, and where I stand personally.

What is AI music, and what isn’t?

From the outset, I think it’s important to understand what AI music is compared to what it isn’t.

Through my love of music, the courses I studied, the practical work and making music myself, one thing has always been clear to me. Someone was always trying to get a step ahead in how they shaped sound, chords and vocals. Someone was always chasing a unique sound. Someone was using Auto-Tune. Somebody was always finding new ways to bring sounds together so the music came out the way they wanted.

The point I’m making is that there’s always been modulation, mixing, compression and so on, to the point where the original sound isn’t really the original sound. In a way, it’s been artificially replicated so the desired sound can be produced. A few examples:

Fake rooms (1947 to 1957). The big echo on a lot of classic vocals came from a room that never existed. In 1947 engineer Bill Putnam turned his studio’s bathroom into an echo chamber. In 1957 the EMT 140 arrived: a large steel plate that vibrates to imitate a room. Abbey Road bought four in the first year. (NPR)
Two people, a whole band (1951). Les Paul rigged a tape recorder so he could play along with his last take and record both together, over and over. On “How High the Moon”, he and Mary Ford sound like a guitar section and a choir. It spent nine weeks at number one in the US. (Sound On Sound)
Artificial Double Tracking (1966). Singers used to sing a part twice to sound fuller. The Beatles found that tedious, so Abbey Road engineer Ken Townsend faked the second take with a second tape machine running a few milliseconds behind. “Artificial” is right there in the name. (Abbey Road)
“Strawberry Fields Forever” (1966 to 1967). The flute at the start is a Mellotron, a keyboard that plays tape recordings of a real flautist. The finished song is two different takes, in different keys and tempos, joined by speeding up and slowing down the tapes until they matched. (Mix)
Auto-Tune (1997). Built on maths a former oil-industry engineer had used to read seismic data, it was meant to fix pitch so quietly nobody would hear it. Cher’s “Believe” (1998) turned it right up on purpose and made it a sound of its own. (NPR)
The modern vocal. Most vocals on records today are “comped”: stitched together from the best phrases of lots of takes, then tidied with pitch and timing tools like Melodyne. (Sound On Sound)
AI before the AI boom (2014 to 2023). LANDR was mastering tracks with algorithms in 2014, and iZotope’s assistants have used machine learning to suggest mix and mastering settings since 2016. For the Beatles’ “Now and Then” (2023), machine learning separated John Lennon’s voice from an old cassette demo. That song won Best Rock Performance at the 2025 Grammys. (iZotope, Grammy.com)

None of these were called AI, but every one of them shaped or replaced the “original” sound. Today they’re just part of how records are made.

Why the distrust now?

For me, AI doesn’t mean the slop: the boom of mass-generated tracks from people taking advantage of it to make money. That seems to be where the discontent comes from, and the distrust of whether the music is authentic.

The numbers back that up:

The flood. On its busiest days in June 2026, fully AI-generated tracks passed half of all new uploads to Deezer, around 90,000 a day. Yet they make up only 1 to 3% of what people actually stream there, and up to 85% of those streams in 2025 were fraudulent. (Deezer)
Most people can’t tell. In a blind test of 9,000 listeners in eight countries, including the UK, 97% couldn’t tell AI music from human music. 80% want fully AI tracks clearly labelled. (Deezer and Ipsos)
UK fans want labels and consent. 81.5% say music made only by AI should be clearly labelled, and 78.5% say AI shouldn’t use artists’ music or voices without permission. (BPI)

The distrust isn’t new either. In 1930, after sound films put cinema orchestras out of work, the American Federation of Musicians ran newspaper ads starring a sinister robot that stood for recorded music (Smithsonian). In 1982 a London branch of the Musicians’ Union voted to ban synthesisers (MU history). In 2010 Time magazine named Auto-Tune one of the 50 worst inventions, and in 2023 its inventor was given a Technical Grammy (Grammy.com).

From my point of view, the artificial side of music has always been there in some form. The problem now is that it’s become far more accessible to create music. So I think the real question is how much of it was AI generated:

Was the music itself, the instrumentation, done completely by AI? Is that acceptable or not?
How much human input went into the style, the adaptation and the composition?
The lyrics: did someone just press “make me a lyric about this kind of song”, and that’s what you’re hearing? No human feeling, thought or process behind it, and no amendments or adjustments?
Is it acceptable that it’s generated and reproduced? Or are some AI-generated songs acceptable and others not?

Even the Grammys draw the line there. Their rules say only humans can be nominated or win, but a song with AI elements can still qualify if the human part is “meaningful and more than de minimis” (Grammy rulebook).

What the music industry is doing

In two years the industry has gone from suing AI companies to setting rules for them:

From lawsuits to licences. In 2024 the major labels sued the two biggest AI song generators, Suno and Udio. Since then Universal and Warner have settled with Udio, and Warner with Suno. In September 2026 Suno launched new models built with Warner, BMG and Believe, while Universal and Sony are still suing it. (Music Business Worldwide)
Streaming services are labelling and filtering. Spotify has pulled more than 50 million “AI slop and spam” tracks out of its recommendations in 2026 (Music Business Worldwide). Deezer tags AI albums and is adding a switch to hide fully AI music from search. YouTube Music started labelling songs made mainly with AI in October 2026 (YouTube). Bandcamp has banned music made wholly or mostly by AI (DJ Mag).
The charts are drawing a line. New global chart rules only count AI-assisted tracks if they’re “substantially human made” and the AI use is disclosed (IFPI). Australia has banned wholly AI recordings from its charts altogether.
Protecting artists’ voices. Sony says it has asked platforms to remove more than 135,000 AI deepfakes of its artists. (Music Business Worldwide)
The UK government backed off. In 2025 more than 1,000 musicians, including Kate Bush and Damon Albarn, released a silent album protesting plans to let AI companies train on music unless artists opted out (Music Business Worldwide). 81% of the 11,520 responses to the government’s consultation said a licence should always be needed. In March 2026 the government said the opt-out was no longer its preferred way forward, but there’s still no new law. (UK government report)

I support the music industry trying to protect the creativity of artists and stop AI just duplicating other people’s work. I think it’s really important that they do.

Where I stand

I’m extremely grateful for what this has opened up for me. As someone who wanted a career in the music industry and started to find my footing in that field, getting access to vocalists and musicians, even for just one track, was a real problem. You had to hire out recording studios and bring people together, and they’d want to be paid. Then you had to actually create the sound. It could be a hell of a lot of work before you could even put a demo together. It never felt that accessible to me.

What AI has done is let me rediscover my old spark: my love of creating sound, music and styles in a way I’ve never been able to before.

Yes, I use AI to generate the music. It lets me bring everything together and produce, in effect, lots of demos of my work and my style. Then I take it into Logic Pro, where I refine and amend the sounds. The lyrics and the composition are mine, and that’s where the authenticity stays. The performances, clearly, I can’t produce myself, because I just don’t have the means. So AI gives me a platform to showcase the music I’d want to make. And what I hope I can show is that it’s really a chance to demonstrate what I can do as a songwriter and sound engineer.

Where I’m conflicted

Where I’m a bit conflicted is this. Even though I’m using AI to showcase my music, I think I’m in line with the music industry: I’m probably against fully generated AI music becoming mainstream. For creativity, songwriting and sound production, I think there should be some substance and authenticity behind it.

Why I’m not doing this for money

The music I’m putting out is there to showcase what I do. That’s why I haven’t reached out to a publisher. I don’t want to publish it. This comes from passion. I love music, and I love making it. For me it’s a documentation of everything I love to do, and that’s enough.

I have no motive around money or monetisation, or putting music out there to generate income. I think that’s probably the wrong focus, and it’s probably what leads to AI slop, which we’re all battling against right now while the industry tries to find its footing.

The best case

If anything ever did come of it, this is how I’d see the best case:

Musicians, artists and vocalists getting in touch to use songs from my catalogue for their own music.
A production company seeing the quality of the work and asking me to write music for their artists.

Either of those would sit right with me.

Where that leaves me

So that’s where I stand: how AI lets me do what I’m doing now, where my moral compass sits on AI in mainstream music, and how I’d want my music used, if that ever becomes a possibility.

I’d like to know where you draw the line. Lyrics? Melody? The performance? Tell me on X (@LoneHeroMusic) or in the comments on YouTube.