Spotify pulls more than 50 million AI slop tracks from its recommendations

Streaming services were built on a simple promise: press play and the app finds the next good song for you. Generative AI has made that promise much harder to keep. Anyone can now produce thousands of passable tracks in an afternoon, upload them through a distributor and hope that a few of them slip into playlists and collect royalties. On 20 October 2026 Spotify published a one year update on how it is fighting back, and the headline number is large: more than 50 million tracks identified as AI slop or spam were taken out of its recommendations this year alone.

The announcement, titled "One Year In: Putting Human Artistry First on Spotify", appeared on the Spotify Newsroom and was quickly picked up by The Verge, NBC News, Variety and Engadget. This article explains what Spotify actually changed, how its detection works, which numbers are new and which are not, and why the update leaves some obvious questions open.

What Spotify announced

The core claim is about recommendations, not deletions. Spotify says that in 2026 it has taken more than 50 million slop and spam tracks out of the systems that suggest music to listeners. Those systems include algorithmic features such as personalised mixes, radio and autoplay, and the editorial playlists curated by Spotify staff. According to the company, its editors now apply the same standard and do not program content that has been identified as slop or spam.

The work started with listener behaviour. Earlier this year Spotify looked closely at the tracks people were skipping, ignoring or explicitly telling the app they did not want to hear. A clear pattern emerged, the company says: those tracks were consistently the ones its systems had identified as slop or spam. Over the past six months Spotify then made what it describes as numerous updates to its recommendation system to push that content out and surface more music by authentic human artists. Since then, it reports a marked decrease in listener complaints about AI slop showing up in recommendations.

The second number in the post is older. Spotify repeats that it has removed more than 75 million spammy tracks from the platform altogether for serious or repeated violations of its rules. That is the same figure the company gave in September 2025, when it first presented its AI music policy. Engadget pointed this out, and a Spotify spokesperson told Music Business Worldwide that deletions have continued, without giving an updated total.

Removed from recommendations is not removed from Spotify

The difference between the two numbers matters. A track that is taken out of recommendations can still exist on Spotify. You can still find it by searching, open it from an external link or hear it if you follow the artist. It simply stops being pushed into your Discover Weekly, your radio stations or the editorial playlists that drive most streams for unknown acts.

From a business point of view this is a softer and cheaper intervention than deletion. Removing a track requires a confident judgement that it breaks the rules, and mistakes can hit real musicians who happen to use AI tools. Demoting a track only changes what the algorithm chooses to show. For most slop the effect is similar, because a song nobody is steered towards earns almost nothing. But it also means that Spotify's library still contains a large amount of material that the company itself regards as low quality.

How the detection works

Spotify is open about the fact that there is no single test for AI slop. In its post the company says it combines several signals:

• Listener engagement and feedback. Skips, ignored tracks and explicit "don't want to hear this" actions.
• Upload patterns. Accounts that push out unusual volumes of tracks in a short time.
• Metadata. Information delivered with each release, such as credits, titles and genre tags.
• Third party AI detection tools. Several external classifiers that estimate whether audio was generated.

The company also lists the tactics it is actively hunting: mass uploads, duplicate tracks, search engine style manipulation of titles and tags, networks of spammy playlists and artificially short tracks designed to farm plays. Very short tracks are an old trick, because royalties on most streaming services are counted per stream after a short minimum listening time.

Notably, Spotify admits the limits of the technology. "AI detection technology in particular has limitations, and it will only get harder as generative AI evolves," the post says. That is an honest statement. Audio classifiers can be fooled by small edits, and every new generation of music models removes artefacts that older detectors relied on. Combining many weak signals is a sensible answer, but it also means the system is a probability estimate, not proof.

The other tools from the past year

The 50 million figure is the headline, but the post also summarises a set of features Spotify has rolled out since its first AI policy announcement:

• AI Credits. Launched in April 2026. Artists now submit tens of thousands of credits every day that show which roles or instruments in a song were made with AI. Listeners can see them in the song credits in the Now Playing view on mobile.
• Verified by Spotify. More than 300,000 artists are verified. Spotify says they account for more than 99 percent of what listeners search for, and more than 70 percent of them are independent.
• AI Persona badges. Thousands of artist profiles whose identity is believed to be AI generated now carry a label. These personas are excluded from editorial and algorithmic recommendations by default, unless a listener follows them. Spotify says it is the only major streaming service that labels AI generated artist identities.
• Artist Profile Protection. A beta that lets musicians review releases that list them as an artist before they appear on their profile. Thousands of artists have opted in, and the tool will be available to all artists by the end of the year.
• The 1,000 stream threshold. Since 2024 tracks need at least 1,000 streams in a year before they earn royalties. Spotify says tracks with strong indicators of being AI generated are ten times more likely to fall below that line.

The threshold is the least visible of these measures and probably one of the most effective. Slop farms rely on volume: thousands of tracks that each collect a small number of plays. A minimum of 1,000 streams per track cuts off exactly that long tail. Spotify did not publish how many tracks fall below the threshold in total, so the ten times figure is a relative comparison, not an absolute count.

Why AI slop is a money problem for musicians

For listeners, slop is mostly an annoyance. For artists it is a direct financial issue. Spotify, like most streaming services, pays rights holders from a shared pool based on their share of total streams. NBC News quoted Morgan Hayduk, chief executive of the fraud detection company Beatdapp, who described the mechanism plainly: if the denominator grows because of AI slop, the percentage that each legitimate artist or rights holder receives shrinks.

Engadget gave some examples of how much attention suspected AI acts can attract. Artists widely believed to be synthetic, such as Sienna Rose and The Velvet Sundown, have more than 100,000 monthly listeners on the platform. Another act, Breaking Rust, topped a Billboard digital downloads chart last year and has around 700,000 monthly listeners. According to a report from July cited by Engadget, the ten most listened to AI artists on Spotify had earned about $2.4 million. Engadget's later update notes that these profiles now carry the AI Persona label.

Not every use of AI is slop, and Spotify is careful to make that distinction. Many musicians use AI for stems, mastering, vocal effects or ideas, and the AI Credits system is designed to let them say so openly. The target of the new measures is mass produced content and impersonation, not the use of new tools as such.

How other services handle AI music

Spotify is not the strictest service in the market. The Verge and Engadget listed how competitors approach the problem:

• Deezer said this month it will let users filter out AI generated audio entirely by the end of the year.
• YouTube Music puts labels on AI songs.
• Bandcamp largely bans AI generated music.
• Tidal does not pay royalties for tracks that are fully AI generated.

Spotify has said earlier this year that its priority is harmful uses such as spam and impersonation, rather than filtering music based on how it was made. That position explains why there is still no switch in the app that hides all AI generated music, even though Spotify's own data shows that many listeners skip it.

The criticism

Engadget's headline accused Spotify of "dancing around" its AI problem, and the criticism has three parts. First, the 50 million tracks were demoted, not necessarily deleted, so the catalogue keeps growing with content the company itself considers worthless. Second, the 75 million deletion figure has not moved in more than a year, which makes it hard to judge how much enforcement has actually increased. Third, Spotify is pushing its own generative features at the same time. The Verge noted that it is partnering with Universal Music Group and Believe on a tool for AI remixes, and Engadget pointed out that these remix and cover features are announced but not yet available.

There is also a transparency gap. Spotify did not say how many tracks are uploaded each day, what share of them it flags, how many human tracks were wrongly demoted or how listeners can appeal. Without those numbers, 50 million is impressive but impossible to put in context. It could be a large share of new uploads or a small one.

Our take

The most interesting detail in Spotify's post is not the 50 million figure. It is the admission that listeners found the problem first. Skips and "don't want to hear this" signals pointed the company to slop more reliably than any single AI detector. That is a useful lesson for every platform dealing with generated content, from music to video to news: behaviour data at scale is often better than a classifier that tries to guess how something was made.

At the same time, demotion is a half measure. It protects the default listening experience, which is what most users touch, but it keeps the economic incentive alive for anyone who can get a few tracks past the filters. Royalty thresholds, verification and profile protection attack that incentive more directly, and they deserve more attention than the headline number.

The real test will come at the end of the year. If Artist Profile Protection reaches every musician on schedule and Spotify starts publishing regular figures on uploads, flags and appeals, its "human artistry first" message will be easier to believe. If Deezer's filter turns out to be popular, Spotify may face pressure to offer listeners the same choice. For now, the update shows that AI slop is no longer a fringe problem for streaming. It is part of the core product work.

Sources: Spotify Newsroom, The Verge, NBC News, Variety, Engadget.