Every morning, millions of people open Spotify, Apple Music, or YouTube Music and let an algorithm decide what they hear next. A playlist auto-populates. A “recommended for you” row fills with artists who feel oddly familiar. A new song by an independent artist from Lagos or Atlanta either makes the cut or quietly disappears into the digital void. The question of who gets heard in the streaming era is no longer just a matter of talent, timing, or label backing. Increasingly, it is a question of data.
The Music That Built the Industry
To understand who algorithms privilege, it helps to remember who built the music that the entire industry stands on. Jazz, blues, rock and roll, R&B, hip-hop, soul, funk, reggae, afrobeats, these genres were created and developed primarily by Black artists, many of whom were systematically denied the credit, ownership, and economic rewards their work deserved. Radio programmers in the mid-twentieth century refused to play “race music.” Labels signed Black artists to exploitative contracts. White artists covered Black musicians’ songs and outsold the originals. The industry’s wealth was built substantially on Black creativity while directing that wealth elsewhere.
Streaming promised to change the equation. No more radio gatekeepers, no more label-controlled distribution pipelines, no more payola. A producer in Accra or a rapper in Chicago’s South Side should, in theory, have the same shot as any signed act. In practice, the algorithms have inherited many of the same structural inequities, just re-encoded in machine learning systems rather than corporate policies.
The Invisible Curator
Algorithms are, at their core, pattern-recognition machines. They learn from what users do, what they skip, replay, save, share, and let run to the end. Spotify’s recommendation engine draws on hundreds of data signals per listening session. It cross-references your taste with millions of listeners who behave similarly. It analyses the audio characteristics of songs, tempo, key, energy, danceability, through a system called audio feature extraction. Then it builds a picture of you, and serves music it predicts you will like.
In theory, this is a breakthrough for discovery. A bedroom producer in Lagos should have the same shot as a signed act from Los Angeles. In practice, it is not quite that simple.
The Feedback Loop Problem
Algorithms amplify what already works. A song that gets early streams, because an artist has an existing fanbase, or because a label paid for playlist placement, signals to the algorithm that it is worth pushing further. More streams follow. The algorithm learns to recommend it more. The rich get richer.
This creates what researchers call a popularity feedback loop. A 2022 study published in Nature Human Behaviour found that algorithmic recommendation systems consistently favour already-popular content, meaning a small percentage of artists capture a disproportionate share of listening time, even on platforms that claim to democratise music discovery.
For Black independent artists, this creates a particular bind. Major labels, who have the promotional budgets to generate early streaming momentum, have historically signed fewer Black artists in proportion to their contribution to the industry’s overall revenue, and have offered those artists less favourable terms. The capital asymmetry that shaped the pre-digital industry now shapes the algorithmic one.
Genre Bias and the Miscategorisation Problem
The bias in recommendation systems is not always intentional, but it is real and measurable. Because these systems are trained on historical data, they tend to reflect historical inequalities. Black music genres have been especially vulnerable to miscategorisation.
Afrobeats, amapiano, dancehall, Afro-soul, and other Global South genres built massive organic audiences in the 2010s. But algorithmic infrastructure was largely built by engineers trained in Western pop structures, who often miscategorised these genres or grouped them under catch-all labels like “world music”, a designation that functioned less as a descriptor and more as a marginalisation. An amapiano artist from Johannesburg and a cumbia producer from Medellín would both be filed in a bucket that told the algorithm: niche, international, low-mainstream-appeal. The recommendation engine learned to treat them as exceptions rather than as mainstream music with global reach.
A 2021 report by the USC Annenberg Inclusion Initiative found that artists of colour received fewer algorithmic recommendations than white counterparts with comparable listener engagement metrics. The system was not neutral, it was encoding the industry’s existing hierarchies at scale.
Playlist Economics and Black Music
Alongside organic algorithmic discovery, a secondary economy has emerged: the business of getting onto playlists. Spotify’s editorial playlists, Rap Caviar, Today’s Top Hits, RapCaviar specifically, are curated by human editors and can be career-defining. Rap Caviar, one of Spotify’s most influential playlists, focused primarily on hip-hop and has been credited with breaking major Black artists. But the playlist economy also has a shadow side.
Third-party services that offer “playlist placement” for a fee have proliferated, some of them fraudulent, pushing music to bot accounts or low-engagement listeners. Artists with promotional budgets, typically those already backed by major labels, can invest in these services. Independent Black artists without that infrastructure are locked out, regardless of the quality of their work.
There is a longer history here worth naming. The practice of paying for radio airplay, known as payola, was made illegal in the United States in 1960, in part because it was being used to suppress Black music at a moment when artists like Chuck Berry and Little Richard were crossing over to white audiences. The algorithmic equivalent of payola is more opaque, harder to regulate, and no less consequential.
What Good Data Could Look Like
None of this means algorithms are inherently bad tools. The question is what we want them to optimise for. Currently, most streaming algorithms are optimised for engagement, keeping users on the platform as long as possible. But engagement can be gamed, and it reflects existing taste rather than expanding it.
Some platforms are experimenting with alternatives. Deezer has tested an “artist-centric” royalty model that weighs active listening more heavily. SoundCloud has piloted direct fan-to-artist payment flows. There are proposals for “diversity bonuses”, algorithmic adjustments that reward platforms for recommending music from underrepresented artists.
Human curators still matter, too. Platforms like Bandcamp, which rely on editorial curation and community discovery rather than pure algorithmic recommendation, have consistently surfaced a more diverse range of artists. Tidal, co-owned by a group of Black artists including Jay-Z, Beyoncé, and Rihanna, built part of its mission around fairer compensation and better promotion of Black music. The lesson is not to abandon data, but to be intentional about what the data is asked to find.
The Bigger Picture
Music has always been political. Who gets a record deal, who gets airplay, who gets reviewed, these decisions have never been purely meritocratic. Black artists have navigated gatekeeping at every stage of the music industry’s history, from the segregated charts of the 1950s to the streaming dashboards of today. The digital era promised to change that, and in some ways it has: a rapper in Houston or a singer-songwriter in Nairobi can reach a global audience without a major label. But algorithms have inherited many of the same biases that shaped the industry before them, and they operate at a scale and speed that makes them harder to challenge.
Understanding how these systems work, who builds them, what they are optimised for, and whose tastes they encode, is the first step toward demanding something better. The next great voice in music is out there. Whether the algorithm finds her is a choice, not a foregone conclusion.
