Last week, I spoke about AI on a panel on the future of music at Valencia Digital Summit. The week before, I was one of the invited judges for the 2026 AI Song Contest, reviewing both the music and the process documentation submitted by its creators. So I’ve spent a considerable amount of time recently analysing what happens when almost anyone can produce a complete, professional-sounding piece of music… and hearing the reactions when I talk about this change.
I’d say I’m a cautious optimist for creators and creativity. And much of what I’ve learned over these two weeks has reinforced why.
AI has made it extraordinarily easy to produce something that sounds like a song. But it hasn’t necessarily made it any easier to understand what makes a song “good”, or to create something we’d want to listen to more than once.
We tend to think of creative barriers in terms of access to tools, creative education, money, or technical expertise. Remove those barriers, we assume, and more people will be able to create. Generative AI appears to offer exactly that opportunity in music, giving us an ideal test case for whether this assumption is actually true.
And this is where the idea of Creative Privilege becomes particularly interesting. The barrier to producing something that sounds “professional” has almost disappeared. But the barriers to communicating something through music that genuinely touches others, and turning an idea into something another human might cherish, seem to remain.
I’ve debated long and hard about definitions of art with my architect partner recently, so we’ll leave that for another post, but let’s use the term loosely here. Let’s call a piece of music that moves us, that we want to playlist to listen again, a work of art.
If access to professional-level tools and composition aids is no longer the primary barrier to making this kind of art, what is? And can we ever take away those barriers with AI?
Spoiler: I don’t think so.
When everything sounds like music
One of the strangest things about judging an AI song contest is just how familiar so many of the submitted songs sound. You’ll hear polished pop ballads, soulful R&B arrangements, synthesised orchestras and convincing vocal performances. The production quality is remarkable, particularly considering how quickly these systems can generate a finished track, and with limited iterations.
But very quickly, very many of the submissions fade into one. It all sounds like music we’ve already heard. Not necessarily a particular song, but a recognisable collection of musical conventions. There’s nothing remarkable that makes me want to listen again. There’s nothing that makes me rewind a track and want to bathe in it for a few moments more.
The models are extremely good at reproducing the characteristics of established musical styles. What they’re less consistently good at producing is a reason for those characteristics to exist.
A chorus becomes dramatic because that’s what choruses in this genre tend to do. A lyric expresses heartbreak because heartbreak is part of the stylistic vocabulary. An arrangement progresses through familiar sections without necessarily developing a musical idea.
The result is technically convincing, but it doesn’t tell me anything more about humanity, or the singular humans behind the track, than a box of Pop-Tarts. And so, my all-too-human brain finds it boring. Aspects can be beautiful, but overall it’s flat, without personality, inhuman.
The songs I responded to the most in the submissions I judged were those that let humanity shine through as a defiant subjugation of AI as just another tool. Unprocessed voice recordings, intentionally left almost shockingly raw to juxtapose against the all-too-perfect orchestration, or a forced restriction, for example using just a handful of AI-generated samples, repeated, folded and transformed to complete an entire exploratory track. These stood out in an otherwise bland box of chocolates.
My family and I laugh at the Lindt Lindor chocolate adverts, which pretend each chocolate is carefully made by hand, whereas everyone knows they’re churned out on a production line. They are too perfect, and therefore can be bought cheaply. Real handmade truffles come with a justified price tag. We will pay for that quality.
But is it the fact that something has been made by hand that gives it value? Or is it the creative judgement of the person making it, the decisions they’ve made about what to include, what to leave out, and what makes the finished thing worth coming back to?
Because that’s where I think the distinction becomes interesting. The problem isn’t that AI is making music. It’s that making music which sounds professional has become so easy that we risk confusing the quality of the production with the quality of the creative decisions behind it.
Intentionality was the difference between generating and creating
One of the things I particularly enjoyed about judging the AI Song Contest was that we weren’t just asked to assess the finished songs. We were also given process documents, explaining how each creator worked with AI, the decisions they made, and how their ideas developed. Reading these alongside listening to the songs was pretty revealing.
Some projects demonstrated enormous technical effort to explore genuinely interesting ideas, yet resulted in music that was difficult to listen to. Others sounded perfectly polished, but offered very little beyond the conventions of a standard generative output.
And neither outcome necessarily meant that the creators lacked creative ambition.
What seemed to make the difference was what happened between having an idea and recognising whether the finished work actually expressed it.
Intentionality is something I return to constantly in my teaching at Berklee College of Music, Valencia Campus. In fact, there’s hardly a practical lesson where I don’t emphasise it. For me, intentionality means being clear about what you, as a creator, are trying to do, express or test at any particular stage of your creative practice, and then using that intention to evaluate whether the decisions you’re making, and the work you’re producing, are actually moving you towards that goal.
Without that intentionality, it’s all too easy to get lost in the labyrinth of creative or entrepreneurial practice, or to start listening to the opinions of others when they have no actual idea of what you set out to do. Of course, every listener or audience will bring their own interpretations and intentions to whatever work you place in front of them, and that’s out of your control. But your own intentionality should be your north star throughout the production process.
And this is exactly what distinguished the strongest submissions. They weren’t necessarily the most technically sophisticated, or the most polished. They were the ones where the human creator had made decisions that fundamentally changed the direction of the work, guided by a clear intention of what they wanted to achieve.
As mentioned above, those two standout examples made this particularly clear. Removing the synthetic voice and replacing it with an almost shockingly imperfect human performance was an intentional decision to preserve emotional authenticity. Restricting the available musical material so severely that the creator was forced to find entirely new ways of working with it was an intentional decision to explore what could emerge from those constraints.
In both cases, the creative achievement wasn’t simply what the AI generated. It was what the human intended to do with it, and why.
(Incidentally, in my work testing AI production systems for video games, we’re finding that intentionality is one of the major blockers to wholly agentic systems producing compelling creative outcomes. But that’s another future post. So much to say!)
When the music sounds finished before the creativity is
I’ve written previously about creative confidence as something that develops through experience, rather than something we need to possess before beginning creative work.
We develop it by making things, experimenting, receiving feedback, revising ideas, and gradually learning to recognise what works and what doesn’t. We learn when to introduce something unexpected, when to simplify, and when to abandon an idea that isn’t going anywhere.
And through that process, we hopefully learn to trust our own intentions and our own judgement.
But generative AI introduces something shockingly new into this creative mix. It gives us something that sounds finished almost immediately.
Pre-AI, a talented musician might spend years learning to play an instrument, compose, arrange or produce before they could make a recording that sounded professionally finished. Throughout that process, they were developing the critical skills needed to set intentions and evaluate their own work against them. Today, AI reverses that process. The polished recording comes first, before the creator has necessarily developed the ability to understand why it works, or why it doesn’t.
When something sounds convincingly professional, how do you develop the skill to recognise that it’s musically uninteresting? How do you know that the dramatic chorus hasn’t earned its place, or that the arrangement needs something unexpected? How do you know when the lyrics are telling you absolutely nothing?
Perhaps more importantly, if the output sounds good enough, what encourages you to keep experimenting?
This is where Creative Privilege becomes particularly relevant. Because creative judgement is learned through experience, experimentation, knowledge, feedback and, often, other people who help us recognise what we haven’t yet learned to see or hear.
AI may be removing many of the technical barriers to producing creative work, but it doesn’t automatically remove the barriers to developing the judgement and creative processes needed to make that work meaningful.
The creative work is still ours
So, I remain a cautious optimist about AI and creativity.
Not because the technology can now produce a convincing song in seconds, but because of what becomes possible when people use it as material for their own creative exploration.
The submissions I enjoyed most weren’t those that demonstrated how closely AI could reproduce the music we already know. They were those where I could hear something of the person behind the work. Their decisions, their curiosity, their willingness to experiment, and occasionally their refusal to accept the apparently perfect output the machine offered them.
And I think that’s an important distinction in the increasingly enthusiastic conversation about AI democratising creativity. We can remove the technical skills that once stood between having an idea and hearing it realised, but we can’t assume that removing those barriers gives people the experience, confidence and creative judgement needed to make something worthwhile.
What excites me most about AI isn’t its ability to generate a finished creative output, but what it might allow us to explore and discover about our own creative practice along the way. It opens up ideas we couldn’t previously test, possibilities we couldn’t afford to experiment with, and creative directions we might never have discovered without it.
But we still have to know what we’re looking for, recognise when we’ve found something interesting, and have the curiosity to keep going when the first result looks good enough.
Making something another human wants to listen to again is still, and will always be, one of the major challenges of creating music. AI doesn’t change that.
