Music’s Latest Take on AI

Last week I was at Music Ally Connect in London- and as you can well imagine AI took it’s usual place as the topic du jour. Despite being well trodden ground there was some fascinating elements- both trends from the day at large and specific comments that gave insight to how the Music industry is currently both thinking about and preparing for (/guarding against) AI.

Mostly these broadly fell into 3 camps- intense negativity, bullish positivity and cautious ambivalence (... of course, it could be argued that most of our approach to life falls into these framings- but that’s a more philosophical discussion for another day).

Not the best photo... but now you at least know the venue's wifi.

AI Will Be the Death of Us

Not a particularly happy one to start on- but there was a consistent thread of ‘AI is the harbringer of the end’. This was particularly brought into focus by a panelist raising the ‘Dead Internet Theory’. 

In short, the theory suggests that a growing proportion of the internet is now created by machines rather than humans- and, more unsettlingly, is increasingly consumed and interacted with by machines too. What was once a gloriously chaotic, inefficient “world of niches” that was undeniably human has slowly been replaced by an ecosystem optimised for scale, automation, and engagement metrics rather than meaning. The humans have been removed.

Bot activity now accounts for well over half of all internet traffic in some reports, with major platforms quietly acknowledging that a meaningful share of posts, comments, videos, playlists, and even “users” are automated. Entire YouTube channels are generated end-to-end by AI, with synthetic voices, stock visuals, and algorithm-optimised scripts- some of them sitting comfortably among the platform’s most-watched accounts. Reddit has confirmed that AI-generated users are increasingly indistinguishable from humans, while moderation teams fight a losing battle against synthetic engagement. Human attention is still there, but critically it is no longer the primary design constraint.

If this framing is even partially correct, the implications for music are profound. How does an artist cut through when discovery environments are saturated with infinite, frictionless content? How does a label distinguish signal from noise when “engagement” itself may be synthetic?

It’s no accident that, against this backdrop, the industry keeps returning to concepts like live, merchandise, superfans, and direct-to-fan relationships. These are not just revenue lines but are proofs of life- physical presence, financial commitment, and community are harder to fake than streams, views, or follows. In a world where digital interaction risks becoming increasingly hollow, these touch points remain as anchors of authenticity.

In that sense, the fear isn’t really about AI making music. It’s about AI eroding the context in which music has meaning- flattening culture into an endless, self-referential stream of content optimised for nobody in particular.

Whether or not one accepts the full premise of the Dead Internet Theory, it is clearly touching a nerve. And it explains why, for some in the room, AI doesn’t feel like a tool or even a threat to rights, but more of an existential question about what parts of life itself are still human, and how long they will remain so.

AI is only getting better

On a less conceptual plane, the larger companies present tended to frame the AI turn tide moment in a more positive light. Jonathan Dworkin (EVP, Digital Business Development & Strategy, UMG) set out a stall that AI is the best iterative tool ever created, with Laurent Hubert (CEO, Kobalt) noting AI will overall likely ‘grow the pie’ rather than shift or redistribute it. Both were also at pains to state how crucial it is having artists on board with the process- i.e. opt in rather than opt out.

The takeaway here is that perhaps, finally, the industry is learning from the lessons of the past- looking to work with new technologies and the winds of the market direction rather than in opposition. The learning also couldn’t be clearer- those that do engage are likely to receive outsize benefit from growth of the overall pie, which is also what we’ve been seeing with the licensing of generative AI companies.

Stuart Dredge brilliantly presented what he described as the ‘humanisation of AI music’. The technical quality in AI generated music has now come so far as to make it both creatively satisfying and commercially palatable- yet it is the underlying human thread that still connects with listeners. He cited Xania Monet as the first fruits in this evolution- an AI artist created and operated by Telisha “Nikki” Jones, who also 100% authors Xania’s lyrics. It is this ability to scale oneself with AI that is appealing- keeping enough authenticity as to draw fans- Xania’s songs having hit 17 million streams in the two months, including charting on Billboard.

Bandcamp’s latest position is to “Keep Human”, but as The Wire recently explored this is a continuum rather than a binary status. This was framed by Soundexchange’s CEO Michael Huppe at Music Ally as the ‘unanswerable question’-  "How much human input is required to make something copyrightable?". Creators, as ever, are pushing the limits to find out- the industry (again, as ever) needs to keep up. No piece of music is truly fully generated by a computer- it is prompted, curated, edited and refined by a human. In The Wire’s words “authorship becomes legible through constraint design and editorial decision making”.

To what extent does this make something ‘AI’ music? That is what the industry currently finds ‘unanswerable’.

Is this business as usual?

In a sense then, this is nothing new for the music industry. We’ve been through technological disruption before, and are now but facing a similar wave. This time, however, we’re better equipped to successfully ride it- to grow value for all while protecting creators.

This was a wider sentiment. That with caution, AI will generate more overall value for the entire ecosystem, much like previous technological disruptions- the internet, P2P and streaming. Of course it will look different, but the underlying mechanics actually aren’t subject to disruption as they have been previously. This is a volume change- both for better and worse. In some verticals the status quo will change significantly (production music, low-value micro sync), and others simply won’t be touched. In the words of Mark Mulligan - "Niche will be the new mainstream, mainstream will be the new niche".

One area that all agreed will evolve is the artist-fan relationship, of which we’re likely to see tangibly in 2026. Dworkin (UMG) highlighted a move away from the "portfolio business of streaming" to increased depth- including products that promote fandom. It’s the same with Xania Monet- there is no shortcut for the connection between artist and fan, and that usually only comes from a ‘human’ connection.

Social Media scaled that connectivity as never seen before- AI is already being applied in a similar way (including across the same social media channels). At some point, a ‘scaled human interaction’ is in fact no longer human at all- think VR concert experiences, or AI Chat Bots that imitate famous figures. In this new paradigm authenticity will shine through all the brighter- and if AI is smartly applied it will bring new engagement, and subsequently revenue.

So… where are we?

One of the best quotes from the conference came from Darren Hemmings. However you view it, as he put it- “AI is providing a beautiful moment to reflect on our humanity- and the beauty and brokenness within it”.

It is indeed a consequential time- and like always with these moments there will be winners and losers. We must, as a moral imperative, ensure creators of all shapes and sizes are not the losers- but this doesn’t mean that AI should be wholesale rejected. Otherwise we will end up stymieing the opportunity for growth it can provide.

The industry, it seems, has learnt from some of the lessons of the past- but as another panelist put it- “Good models come before good governance, and we're only just starting to think about asking the right questions”.