Why Human Curation Beats the Algorithm

The algorithm knows what you like. It has watched your listening history, weighed your skips and replays, consulted the patterns of millions of listeners with overlapping taste profiles. It is, in certain narrow ways, uncannily accurate. And yet something has gone wrong.

The Discovery Problem

The promise of streaming was discovery. Unlimited access to the history of recorded music: every genre, every era, every geographic tradition, available instantly. For listeners willing to explore, this should have been an unprecedented golden age of musical curiosity.

What happened instead was the opposite. Studies consistently show that streaming listeners’ musical range has narrowed since the advent of algorithmic recommendation. People listen to more music by the same artists they already know. Genre diversity in listening habits has decreased. The algorithm, optimised for engagement rather than expansion, pushes listeners deeper into what they already like rather than out toward what they might love.

This is not a failure of technology. It is a consequence of the specific optimisation target. If you want to maximise listening time, the safest bet is to give people more of what they are already listening to. Discovery is risky — the new thing might not land, and the user disengages. The algorithm is doing exactly what it was designed to do. The problem is with what it was designed to do.

What Human Curation Does Differently

A human curator — a record shop employee, a fanzine writer, a DJ, a magazine editor — brings something the algorithm cannot replicate: judgment informed by knowledge, enthusiasm, and the willingness to say something is better than something else.

The algorithm cannot tell you that one record is more important than another, only that it is more popular. It cannot explain why a particular B-side represents a pivotal moment in an artist’s development. It cannot suggest that the record you need to hear is nothing like anything you have listened to before, but that it will nonetheless change how you hear everything.

Human recommendation requires the recommender to have actually listened, to have formed a view, to have sufficient conviction to stake their credibility on the suggestion. That is a fundamentally different relationship than engagement-optimised playlist generation.

The Fanzine Tradition

The history of music criticism is largely a history of people who cared more than was strictly necessary writing about records for audiences of strangers. The fanzine tradition — from the typewritten sheets photocopied in bedroom offices to the zines that shaped early punk, hip-hop, and independent music discourse — operated on the assumption that a passionate voice with a point of view was more valuable than neutral, comprehensive coverage.

That tradition has not disappeared. It has migrated: to independent music blogs, to Substack newsletters, to carefully curated social media accounts. The common thread is a human being who has listened to a great deal of music, formed strong opinions, and found it worth the effort to share those opinions with whoever will read them.

The algorithm cannot do this. It can surface popularity. It cannot surface significance.

Conclusion

This is not an argument for the past, or against technology, or for the proposition that things were better before streaming. They weren’t, in most respects. The access is genuinely extraordinary. But access without curation is noise. The solution is not to abandon algorithmic tools but to supplement them with the one thing they cannot replicate: the judgment of an attentive human who has put in the listening hours.

Frequently Asked Questions

What are the best alternatives to algorithmic music discovery? Independent music publications, community radio stations, local record shops, curated streaming playlists maintained by human editors, and music-focused newsletters all offer discovery that operates differently from pure algorithmic recommendation.

Does editorial music criticism still influence what people listen to? Less directly than it once did, but critics and editors still shape taste in meaningful ways — particularly within specific communities and subcultures. A recommendation from a trusted source with a track record still carries weight that algorithm-generated suggestions typically lack.

Are streaming platforms doing anything to improve discovery? Most major platforms invest in editorial curation alongside algorithmic recommendation. Some have been more successful than others at building editorial voices that listeners trust. The tension between the two approaches — editorial versus algorithmic — remains unresolved at most platforms.

Is the algorithm making music worse by rewarding certain kinds of songs? There is documented evidence that algorithmic optimisation has influenced how records are made — particularly around song length, opening hooks, and the avoidance of quiet introductions. Whether this constitutes “making music worse” depends on what you value, but it is clear that the algorithm has become a factor in creative decision-making in ways it wasn’t a decade ago.