On October 20, Spotify reported on its first year of fighting AI junk: since the start of 2026, more than 50 million tracks the service treats as "slop" and spam have been removed from recommendations. More than 75 million spam tracks were taken off the platform entirely for serious or repeated violations. Two days later, on the quarterly earnings call, the company's co-CEO explained why Spotify expects AI agents to be the ones playing your music.
What exactly was removed
The 50 million tracks are still there: you can find them in search and play them. But Spotify's editors no longer put them in playlists, and the algorithms no longer push them into recommendations. Slop is the industry's word for the cheap stream of generated content made for plays rather than for music.
Spotify doesn't look for a single "this is AI junk" signal. It combines several: how a track is played and skipped, how it was uploaded, what's in the metadata and what third-party AI detectors say. It also watches for mass uploads, duplicates, playlist manipulation and artificially short tracks.
Over the year, a few tools grew around this:
- AI Credits - since April, artists mark which roles and instruments in a track involved AI. The note shows up in the song credits. Tens of thousands of these come in every day.
- AI Persona badge - if an "artist" is fully generated, it gets a label and stays out of recommendations unless you follow it yourself.
- Artist Profile Protection - artists see releases under their name before they appear on their profile. It's in beta now and promised to everyone by the end of the year.
- 1,000-stream threshold - fewer plays in a year and a track earns no royalties. According to Spotify, tracks with a high probability of being AI-generated are 10 times more likely to fall short of it.
Spotify itself says openly that AI detectors make mistakes and that it will only get harder. The company doesn't say how many of the 50 million tracks were made with AI and how many are ordinary spam.
The bet on AI agents
On October 22, Spotify posted its quarterly numbers: 788 million monthly users, 306 million paying subscribers and about €5 billion in revenue. On the investor call, co-CEO Gustav Söderström said Spotify is "in one of the few categories that actually stands to thrive in the agentic age".
His reasoning is simple: people hand agents the things they don't want to do. Music is something you listen to yourself - nobody asks an agent to "summarize a song". So Spotify wants to be the service any agent plays on request. It's already connected to Meta's Muse, ChatGPT and Claude. According to Söderström, listener data doesn't go to partners: the agent sends a request, and Spotify decides what to play.
My take
Spotify isn't fighting neural networks, it's fighting the assembly line that uploads thousands of faceless tracks for pennies in royalties. An agent that plays you real artists is good for the service; a stream-farming agent that loops a generated playlist is not. The catch is that the line is drawn by detectors, and the company admits they make mistakes. That means real musicians will get caught too.
What it means for you
If you make music with AI and put it on streaming services, the rules are now clear. Mark AI use in AI Credits, don't upload tracks in bulk and don't make them artificially short: those are exactly the signals Spotify uses to find slop.
You can try generation in GPTunneL: Suno and other music models work in one window with pay-as-you-go pricing. And if you want to see where the line runs between a tool and a replacement for a musician, we have a breakdown of whether AI will replace musicians and a roundup of AI tools for making music.
Sources
- One Year In: Putting Human Artistry First on Spotify - Spotify's report on protections against AI spam, October 20, 2026
- Spotify Reports Third Quarter 2026 Earnings - quarterly results and Söderström's quote, October 22, 2026
- Spotify and Muse - the integration with Meta's agent, September 23, 2026