I’ve been thinking a lot about AI and music lately, not because I’m a tech person — I’m barely competent with the shop’s point-of-sale system — but because the way people discover music is changing, and that affects everything about my business and the artists I champion.

The question isn’t whether AI will be part of music discovery. It already is. The question is whether it’s going to help or hurt the independent artists and small labels that make this industry worth caring about.

Where We Are Now

Every major streaming platform uses AI-driven recommendation systems. Spotify’s Discover Weekly, Apple Music’s personalised playlists, YouTube’s sidebar recommendations — these are all powered by machine learning algorithms that analyse your listening history and predict what you’ll want to hear next.

For mainstream artists, this system works reasonably well. If you listen to a lot of popular indie rock, you’ll be served more popular indie rock. The algorithm reinforces existing preferences and keeps you within comfortable boundaries.

For underground and genuinely independent artists, the picture is more complicated. Recommendation algorithms favour tracks with high engagement metrics — saves, shares, playlist additions. By definition, underground artists have fewer of these signals, which means the algorithm is less likely to surface them.

This creates a feedback loop. Popular music gets recommended more, which makes it more popular, which gets it recommended even more. Niche music stays niche, which means fewer recommendations, which keeps it niche.

The Human Alternative

This is where record shops, community radio, music publications, and word of mouth remain essential. A recommendation from a knowledgeable person who understands your taste and is willing to push you toward something unexpected is fundamentally different from an algorithmic suggestion based on behavioural data.

When I recommend a record to someone in my shop, I’m drawing on decades of listening, a deep understanding of the customer’s taste, and an intuition for what might expand their horizons. I can recommend something because of how it makes me feel, because of the story behind it, because the pressing is beautiful, because the band is from the next suburb over.

An algorithm can’t do that. Not yet, anyway. And honestly, I hope it never fully can, because the human element of music discovery is part of what makes it meaningful.

The Emerging AI Tools

That said, some AI applications in the music space are genuinely interesting.

AI-powered search and curation tools are getting better at understanding nuance. Instead of “recommend something like Band X,” you can increasingly describe a mood, a feeling, or a context and get surprisingly relevant suggestions. This is useful for casual listening even if it’s no replacement for deep expertise.

AI music analysis tools can now identify sonic characteristics of tracks with remarkable precision, which has potential for connecting listeners with music they’d enjoy based on sound rather than metadata. If a tool can identify that you respond to a particular guitar tone or rhythmic pattern and surface Australian independent releases that share those qualities, that’s potentially helpful.

Inventory and customer insight tools are where I see the most practical near-term value for record stores. Understanding which releases to order deeper on, which customers are likely to want a particular new release, and how to time promotions effectively. Some shops have started working with Team400 to build these kinds of insights, and the results have been promising for smaller operators who can’t afford dedicated data teams.

Working with an AI consultancy can also help labels and shops understand listening patterns that inform which artists deserve vinyl pressings — a decision that currently relies mostly on gut feeling.

The Risk for Independent Artists

The biggest risk I see is that AI-driven discovery further concentrates attention on already-visible artists while making it even harder for genuinely underground music to find an audience.

If the primary way people discover music is through algorithmic recommendation, and those algorithms favour engagement metrics, then the artists who benefit most are the ones who are already succeeding. The garage band in Footscray playing to thirty people on a Tuesday night becomes even harder to discover, even if their music is extraordinary.

This isn’t a new dynamic — mainstream media has always favoured the popular. But the scale and pervasiveness of algorithmic recommendation amplifies it.

What We Can Do

Keep supporting human curation. Community radio, independent music publications, record shop recommendations, and word of mouth remain the most effective discovery channels for independent music. Support them financially and with your attention.

Be intentional about discovery. Don’t let an algorithm be your only path to new music. Go to shows. Browse a record shop without a specific purchase in mind. Listen to a community radio station for an hour. Click on a Bandcamp tag you’ve never explored.

Demand better algorithms. If streaming platforms want to claim they support independent artists, their recommendation systems need to actively surface niche and underground music, not just reflect existing popularity. Algorithmic serendipity should be a design goal, not an accident.

My Position

I’m not anti-technology. I use streaming for discovery and I think AI has real potential to help independent artists reach audiences. But I’m deeply wary of any system that reduces music discovery to a popularity contest, because the most important music is almost never the most popular music.

The record store will always be the antidote to the algorithm. Walk in, tell me what you’re in the mood for, and I’ll put something on the turntable that no machine would have picked. That’s the value proposition, and it’s not going anywhere.