Reading Advice: This is not the best post ever written, but is a mildly interesting tale of enshittification (which is a trend in the modern world we should remind ourselves of from time-to-time).
Disclaimer: Very limited number of examples, and large organisations have been known to experiment on their users so strong conclusions shouldn’t be drawn without wider evidence or a systematic study.
Quality Rating: 5/10
I think the best word I’ve never said in real-life is “enshittification”. It was coined by Cory Doctorow in this 2022 article 1, and his 2025 book “Enshittification: Why Everything Suddenly Got Worse and What to Do About It” carried the name. I think it captures well plenty of aspects of the modern internet, and indeed plenty of modern companies in a short-term-focussed mindset.
Spotify
One of the main ways I listen to music is via Spotify. As I was trying to assemble a playlist based on a compilation CD a family member had passed to me, it struck me as more challenging than I anticipated. The general genre was big band leaders of the 1930s and 1940s, and typically they made multiple recordings of their music. Because I’m a little bit particular, I decided to try and match the precise recordings with those of my CD, for which I had exact track lengths. I was lazy and just searched for {track name} and {big band leader’s name}, not thinking anything curious would happen…
Spotify appears to load search results in sequences of 20, and I would typically find that the first 1-10(ish) would match both the track title and artist, and then those in 11-20 would very often not match, and most came up with a “Lyrics Match” label. I immediately started laughing once I realised this still applied to music where there was no lyrics, and for song titles where no reasonable person would include those words as lyrics. To top it off, occasionally the songs supplied under “Lyrics Match” often had no lyrics themselves 2!
This could be dismissed as a curious quirk where someone decided it was important that the first 20 results contained. So I scrolled down, and found that tracks 21-30 seem to fit the search criteria and 31-40 again seem to be almost totally unrelated. The pattern repeats.
The removal of choice for the user has been a standard path that Twitter and many other social media companies have used, but I think there’s something a bit more sinister and comic-book-villain-evil about specifically refusing to share results that have been requested by the user, making them scroll past information that’s obviously irrelevant. Google can serve us adverts (or AI summaries nowadays) in/around more useful content, and we seem to mostly accept it. But like it or not, Spotify also has a similarly market-dominant position, and to willingly degrade the quality and usefulness of their search functionality by 50% (literally) to me is unacceptable, and has led to me concluding that I will never pay for their services.
Questions
My open question is how are these “lyrics matches” determined when the search clearly doesn’t match many/any lyrics? A bit of brief experimentation typically leads them to having one word in common, maybe “similar-ish” styles to the search, but often the common theme is they are popular. Therefore I can’t help but suspect that it’s been seen by spotify as a way to keep people listening (maximising a metric which probably is not directly useful to the user anymore) by selecting music that they suspect appeals to the searcher, or possibly even diverts them away from the kind of music they’re searching for to more monetisable forms of music (typically shorter tracks, see my post on collider bias as a tool).
Here are some more:
- How much of this can be attributed to Spotify being a near-monopoly?
- Is the intention of Spotify to serve user or advertisers nowadays in case like these?
- Would less useful search results lead to more time spent seeing adverts in the app potentially?
- Can artists/record-companies pay to appear (higher?) on this “lyrics match”?
- Is the same behaviour seen by those who do pay for Spotify?
- How much of this can be attributed to Spotify developers/decision-makers not considering that plenty of music does not contain any lyrics?
- How much of this can be attributed to Spotify developers/decision-makers not considering any changes on historical music that often has several versions?
One question I don’t want to think about, but is interesting, is: “in what new ways can Spotify speed-up/enhance the enshittification process?”
Footnotes
The concept is slightly older, however a big part of getting yourself written into history is finding the correct name for something and giving it traction by talking about it with a platform. I see this reasonably often whenever I’ve studied topics in the history of mathematics that had near-simultaneous discovery or popularisation moments.↩︎
In a strict sense, I suppose Spotify were correct for some of my searches, but the cynic in me suspects more by accident rather than design!↩︎