Is Spotify’s AI ‘killing’ Australian music? What we found from analysing more than 2 million tracks

Last year, former Spotify chief economist Will Page compiled a report for the Australia Institute that concluded music streaming algorithms were “killing” Australian music.

The report found that, between 2021 and 2024, there was a 30% drop in Australian artists in the top 10,000 artists streamed in Australia on platforms such as Spotify, YouTube and Amazon.

“The algorithms of streaming services might recognise language, but they ignore geography, which means local music is not typically recommended to Australian audiences,” Page said.

These claims of reduced visibility resonate with Australian musicians, who are concerned their music may be less favoured than the work of more popular global artists.

We fact-checked these claims in new research commissioned by the Victorian Music Development Office, with a focus on Spotify.

While we didn’t find evidence of Australian music being “killed” by AI, we did find algorithms perpetuating conditions that make it difficult for less-established artists to break onto the scene.

How AI shapes streaming

The objective of streaming platforms is to maximise user engagement. Spotify does this by allowing users to discover new music in various ways, including through manual search and exploration, editorial (human-made) playlists, and AI-recommended playlists.

Algorithms have been criticised for amplifying the influence of superstars – and the corporate interests that support them – while also potentially narrowing listeners’ musical preferences.

Spotify’s AI does have a significant influence on the listening habits of its subscribers. But is this a problem?

For many users, AI-recommended playlists are simply convenient. Instead of intentionally searching for new music, they are happy to be recommended tracks they might like.

At the same time, there are concerns algorithmic bias may benefit certain artists over others.

Our findings

Our research, conducted in February 2025, involved analysing 2.27 million music tracks using Chartmetric’s real-time analytics platform.

Our dataset included 12,333 artists and 5,000 editorial and AI-mediated Spotify playlists from seven English-speaking countries: the United States, the United Kingdom, Australia, New Zealand, Canada, Ireland and Jamaica.

Our findings indicate that AI‑generated Australian playlists heavily rely on global listening patterns. They are also less likely than editorial playlists to surface diverse or regionally specific music. This matches the AI recommendations pattern for the UK market.

AI recommendations accentuate US dominance by reproducing US tastes as global “norms”. Our study showed the composition of AI playlists in all countries is very similar to those of the US.

This suggests the US – a much larger market than Australia and the other countries – generates a music footprint that dictates the global trends.

The AI playlists in our sample drew from only a quarter as many unique tracks as the editorial playlists. This further shows how AI playlists, in general, are more concentrated and less likely to recommend local music.

AI’s tendency to recommend “familiar” music also favoured artists from dominant markets such as the US. In our sample, 77% of the US tracks were produced by “established artists”, representing three Chartmetric categories (legendary, superstar and mainstream).

In contrast, only 22% of Australian tracks were being produced by established artists. The artists behind the other 78% of Australian tracks are less likely to be recommended by AI algorithms.

Filter bubbles

Over time, AI playlists – which are more likely to push established US artists – are fed back to users in a loop. This gives more exposure to already popular artists, and further disadvantages less established ones – leading to a “rich get richer” dynamic.

These conditions make it difficult for up and coming acts to break through Spotify’s recommender systems.

One solution might be for Spotify to tailor its AI algorithm to actively boost less-established artists. But for now, the inner workings of the algorithm remain somewhat hidden.