The Great Spotify Purge: 75 Million AI Songs Gone, and It's Just the Beginning

July, 21 2026

Seventy-five million songs. That number is almost too big to picture. To put it in perspective, Spotify's entire catalog currently sits at around 100 million tracks. So when the company's executive head of global artists, marketing, and policy, Sam Duboff, confirmed that Spotify had quietly removed more than 75 million tracks over the past 12 months, it landed like a seismic event. That's not a rounding error. That's not a cleanup. That is a full-scale purge.

And honestly? It's one of the best things that has happened for working musicians in years.

How Did We Even Get Here?

To understand what Spotify just did, you have to understand how bad the problem had gotten.

For the past several years, AI music generation tools have become increasingly cheap, fast, and easy to use. Platforms like Suno let anyone type a prompt and get back a finished song in seconds. Google's Lyria model, accessible through Gemini, offers similar capabilities to anyone with an internet connection. You don't need to know how to play an instrument. You don't need to know anything about music theory, mixing, or arrangement. You just type something like "upbeat lo-fi study music" and the machine does the rest.

That accessibility sounds harmless enough on the surface. The problem is what happens when people figured out those tracks could be uploaded to Spotify and earn streaming royalties. Almost overnight, a cottage industry of bad actors emerged. Operators were uploading thousands of algorithmically generated tracks under fake artist names, attaching misleading metadata, keeping tracks artificially short to maximize play counts, and engineering them to show up in the kind of ambient and background music playlists that rack up streams without anyone really paying attention. Spotify executives started using a specific term for this content: "AI slop."

Duboff himself acknowledged that roughly 100,000 songs are being uploaded to Spotify's servers every single day. Industry estimates suggest that somewhere around 44 percent of all music being uploaded to streaming platforms is AI-generated. Even if that figure is partially inflated, we're still talking about tens of thousands of machine-made tracks hitting the platform daily. The royalty pool doesn't grow to accommodate that. It just gets divided into smaller and smaller pieces. Every fake ambient piano loop sitting on Spotify is quietly siphoning money away from the artists who are genuinely trying to build careers.

How Spotify Is Actually Detecting This Stuff

One of the most common questions around this story is a fair one: how does Spotify know? If someone uploads an AI-generated track under a fake artist name, how does the platform identify it as spam?

The answer is a combination of detection tools, behavioral analysis, and metadata standards that Spotify has been building and refining over the past couple of years.

The most obvious category is voice cloning. If an uploaded track contains an AI-generated replica of a recognizable artist's voice without authorization, Spotify now removes it immediately, regardless of what the uploader claims about its origins. That's a clear line.

Beyond that, Spotify uses machine learning to detect behavioral patterns that signal fraud. Mass uploads from a single account or distributor, duplicate or near-duplicate content uploaded under different artist names, tracks that appear to be engineered to mimic playlist-friendly content without any evidence of genuine engagement, artificial streaming activity where the play counts don't match any real listener behavior. When those signals stack up, the tracks get flagged, downranked, or removed.

The platform has also adopted the DDEX metadata standard, which allows labels and distributors to tag whether a release is fully AI-generated or just AI-assisted. That distinction matters, because "AI-assisted" can mean something as mundane as using software for mastering, which real artists have been doing for years. Spotify surfaces that information to listeners when it's disclosed, so people can make their own informed choices about what they're listening to. On top of all that, there's a dedicated monitoring team whose entire job is tracking how spam tactics evolve and adapting the detection systems accordingly.

None of this is perfect. The arms race between detection and evasion is real, and as Spotify's filters get sharper, bad actors will find more sophisticated ways to blend minimal human edits into AI-generated output to slip through. But the commitment to building these systems in the first place represents a meaningful shift in how the platform thinks about its responsibilities.

What Spotify Is NOT Doing

This is an important distinction, and one that's worth being clear about.

Spotify is not banning AI music.

The company has been explicit about this. If a creator holds the commercial rights to an AI-generated composition, avoids unauthorized voice cloning, and demonstrates genuine creative intent, their music is still welcome on the platform. Authorized AI remixes, AI-assisted mixing and mastering, experimental work that blends human performance with machine generation: all of that is still allowed.

What Spotify is targeting is specifically fraud. The spam operations, the fake artist farms, the bot-driven royalty manipulation. The distinction matters because a blanket ban on AI music would be both unenforceable and arguably unfair to artists who are legitimately experimenting with new tools. That's a complicated conversation worth having. This purge isn't that conversation. It's something more specific and more necessary.

The Real Cost of "AI Slop"

For anyone outside the music industry, it might be tempting to treat this as a niche problem. A bunch of fake songs cluttering up a platform, who cares?

But the financial reality is serious. The royalty pool on streaming platforms isn't infinite. When fake tracks accumulate millions of plays through artificial means, that money doesn't fall from the sky: it comes directly out of what legitimate artists earn. An independent musician whose song gets 100,000 genuine streams is competing for the same royalty pool as a bot farm running 10 million artificially inflated streams on ambient noise tracks. The fake plays don't create new money. They just dilute the real one.

This has been building for a while. Back in 2023, Spotify pulled tens of thousands of AI-generated songs linked to a platform called Boomy after detecting artificial streaming activity. That was a preview of what was coming. The tools only got cheaper and more powerful from there. The 75 million figure is what happens when the problem is allowed to scale for a few years before decisive action is taken.

Is This Actually a Win for Real Artists?

The answer is yes, with some important caveats.

The most direct benefit is a cleaner royalty pool. If 75 million fraudulent tracks are no longer siphoning streams and payouts from the system, that money flows toward the music people are actually choosing to listen to. For independent and emerging artists especially, even a small shift in how royalty revenue is distributed can matter.

There's also a discovery argument. One of the consistent complaints from both listeners and artists over the past few years is that AI-generated filler content was cluttering playlists and making it harder for real music to surface. Ambient music categories, study playlists, sleep playlists: these had become particularly overrun. Cleaning out that noise creates more room for actual artists to be heard.

The verification piece matters too. Spotify is trying to build systems that help listeners distinguish real from fake. The mass deletion and the verification program are two sides of the same effort: reduce the noise, amplify the signal.

The Caveats Are Real Too

Duboff was candid about the difficulty of the situation. He used the phrase "blurred lines" to describe the challenge of drawing clean distinctions around AI use in music, and that candor is worth taking seriously. Real artists are increasingly using AI tools in their normal workflow, for melody generation, lyric assistance, arrangement ideas, mastering. Where does legitimate creative use end and content spam begin? That line is genuinely hard to draw, and Spotify's own CEO had recently told people to stop calling AI music "slop" at all, which tells you something about how complicated the internal politics are.

There's also legitimate concern that aggressive filtering could catch real music in the crossfire. An independent experimental artist using AI tools in unconventional ways might trip the same pattern-matching signals as a spam farm. These systems are not foolproof, and the stakes of a false positive, an actual artist's music getting deleted, are real.

IFPI and RIAA have been pushing for voluntary labeling programs that would give the industry more structured tools to navigate this. Whether those programs develop into something with real teeth, or stay as soft-touch gestures, remains to be seen.

Is This the Solution?

Probably not the whole solution, but it's a start that matters.

The spam problem in music streaming is fundamentally an economic one. As long as uploading AI-generated content to a streaming platform is a viable way to skim royalties with minimal effort and almost no risk, people will keep doing it. Detection tools help. Removal at scale helps. But the underlying incentive structure won't change until the financial math changes, and that requires ongoing enforcement, better distributor accountability, and possibly structural shifts in how royalties are calculated and distributed.

What Spotify just did is the equivalent of finally cleaning out a filter that had been clogged for years. It's necessary maintenance, and the platform deserves credit for actually doing it at this scale. But the filter will clog again unless the systems around it continue to evolve.

For working musicians, the message here is worth holding onto. The platforms are starting to act. The scale of 75 million removed tracks is proof that there's finally some will behind the enforcement. That doesn't mean the fight is over, but it's a different position to be in than it was two years ago, when most artists were left asking whether anyone was paying attention at all.

Turns out, they were. And they finally did something about it.