Spotify Removes 50 Million AI-Generated Tracks From Recommendations

On September 20, 2025, Spotify announced a suite of measures to protect artists and listeners from what it termed AI-generated spam and "slop." One year later, the company published an official progress report with numbers that reveal both the scale of the problem and the scope of its response. Since the announcement, over 50 million tracks identified as AI-generated have been pulled from the service's recommendation engine, and an additional 75 million have been removed entirely from the platform for serious or repeated violations of Spotify's terms.

What "slop" means

The term "slop" — literally referring to low-grade animal feed — has been adopted across the technology industry to describe content generated at scale by AI with the primary purpose of capturing attention, gaming algorithms, or siphoning royalties from streaming platforms. For Spotify, the most prevalent abuse patterns include:

- AI-generated tracks that impersonate established artists without authorization; - Mass-produced duplicate songs designed to occupy search results; - Manipulative SEO techniques to hijack recommendation systems; - Artificially cloned vocals representing non-existent performers.

Spotify's own investigation into listener behavior revealed a clear pattern: the tracks users skipped, ignored, or reported as unwanted were consistently those the platform's systems had already classified as slop or spam. This confirmed that the problem was not marginal — it was directly affecting the music discovery experience at scale.

How Spotify identifies and removes AI content

According to the company's official blog, its detection system combines multiple signals. Users can flag unwanted tracks, and algorithms analyze upload patterns, inconsistent metadata, anomalous streaming behavior, and information from third-party AI detection tools. The company acknowledges, however, that AI detection technology has inherent limitations — and that the challenge will only grow more difficult as generative models continue to evolve.

Starting in mid-September 2026, Spotify began displaying an "AI Persona" badge on artist profiles whose public identity appears to be artificially generated. The badge appears on the profile banner, in the "About" section, in search results, and on track rows within playlists. Profiles tagged as AI Personas are automatically excluded from all editorial and algorithmic playlists — unless the user explicitly follows them.

The 1,000-stream threshold

Another significant change is the 1,000-stream minimum-earning policy: tracks must reach at least 1,000 streams within a 12-month period to generate royalties. Before this policy, revenue was distributed proportionally across a pool that included millions of inactive tracks — and the company estimates that roughly $47 million per year was redistributed in tiny payments that rarely reached artists because they fell below distributor minimum payout thresholds. Under the new rule, that money is reallocated to eligible tracks, directly benefiting active creators.

Spotify states that 99.5% of all streams on the platform are already from tracks with at least 1,000 annual streams, and that each eligible track earns more under the new policy. Tracks with fewer than 1,000 streams represent only 0.5% of total streams and royalties, but the sum of those monthly fractions amounts to tens of millions of dollars wasted annually on accounts that are never cashed out.

The broader context

Spotify's move does not occur in isolation. Competitors are also taking action: Deezer reported in July 2026 that 50% of daily uploads were classified as AI-generated at June's peak — roughly 90,000 tracks per day — yet AI represents only 1% to 3% of actual listening. Tidal has stopped paying royalties on fully AI-generated tracks. The RIAA and the Grammys have proposed industry-standard labels distinguishing recordings created entirely by AI from those using AI as an auxiliary tool. CISAC and PMP Strategy estimated that approximately $4.6 billion in annual artist revenue could be lost by 2028 if the problem goes unaddressed.

Interestingly, in May 2026 Spotify struck a deal with Universal Music Group to enable fans to create AI-assisted covers and remixes under a "consent, credit, and compensation" model. The company is attempting, therefore, to draw a fine line: blocking fraudulent activity while permitting creative and transparent use.

What to watch

The figures Spotify published — 50 million removed from recommendations and 75 million from the platform — are striking on their own. But the fundamental question remains: how does detection actually work in practice, and how long will it remain capable of keeping pace with advancing audio-generation models? The company openly acknowledges that AI detection technology has limitations and that the challenge will only intensify. Meanwhile, the market continues to transform, and real artists remain trapped trying to stand out amid the noise that AI itself created.

How will the detection ecosystem evolve as generative audio models become indistinguishable from human recordings — and will the industry ever agree on a technical standard that doesn't simply become another vector for gaming?

Sources: Spotify Newsroom, Variety, Music Business Worldwide

✓ Independent sources cross-checked and verified before publishing