Think all songs sound the same? It’s proven – Spotify’s algorithms are killing interesting music

The global recorded music industry generated more than $28 billion in revenue in 2023 – but despite all that cash sloshing through the system, the songs topping streaming charts are sounding increasingly samey.

New research suggests that's no accident, but the predictable outcome of the way algorithms, institutions, and social pressure interact to shape what we listen to.

A new paper by researchers in Italy argues that musical taste isn't a fixed parameter we bring to streaming services. It is rather continuously reshaped by what those services choose to show us. And once algorithmic curation crosses a certain threshold, the diversity of what listeners consume collapses – and quickly.

Modelling the impact

The researchers built an agent-based simulation modelling 200 listeners choosing from 80 songs across four very different musical ecosystems: Italy's Festival di Sanremo, Brazil's regional polyphony, South Korea's K-pop oligopoly, and the United Kingdom's independent scene. Each was calibrated against real-world streaming data drawn from Spotify across 73 countries.

The team found that even a tiny tweak to the algorithm could massively reduce the range of music that listeners hear. Push the strength of algorithmic curation from 0.7 to 0.9 on the researchers' scale, and so-called consumption entropy – defined by the academics as the measure of how evenly listening is distributed across available songs – drops by 23%.

The mechanism isn't simply that algorithms recommend bad music. It's that algorithms create a feedback loop with three moving parts that all reinforce each other.

The Cybernews community is talking about this. Be a part of the conversation.

Recommendation systems learn what listeners already like and serve up more of it, narrowing the range of songs anyone actually encounters. Then songwriters and labels watch the streaming numbers and shift what they produce towards whatever is performing well, so the pool of music being made starts to converge on a smaller and smaller stylistic target.

Then what happens – and this is the part the paper takes most seriously – listeners' brains adapt to the narrowed environment. We come to find the restricted range of music fine.

Changing listening habits

These changing listening habits have a real impact on the music that’s created. Italy's Sanremo festival shows what the producer side of that loop looks like at its most extreme. Just 11 songwriters produced roughly two-thirds of the 30 songs competing in the 2025 edition – prompting consumer association Codacons to file an antitrust complaint over what it called a "discography caste". That’s an issue, the researchers say, but it’s not always the case.

Brazil tells the opposite story – and is the closest thing the paper finds to a healthy ecosystem. Samba, MPB, sertanejo, funk carioca, pisadinha, and dozens of other regional genres each have their own audiences, their own production communities, and their own reward structures. Brazilian streamers love local content – 84% of streams – which means Brazil maintains the highest consumption diversity.

Another country the research looked at was the UK, which showed the consumption side of the feedback loop. Its production ecosystem is genuinely diverse, but British music accounts for only 29% of UK streaming, compared with 55% American content.

To try and stop the homogenisation, the researchers recommend maintaining independent generative sources, building diversity safeguards into recommendation algorithms, and treating publicly funded festivals as ecosystem stewards.

It’s also important to invest in music education, they say. Without such intervention, the paper warns, the algorithmic feedback loop will keep narrowing what's possible to hear – until listeners' brains become "very good at predicting a very narrow range of musical experiences".

Unlock more exclusive Cybernews content on YouTube.