The streaming economy has always been built on a quiet premise: that music is scarce enough to be sorted, priced, and fairly paid. That premise is evaporating.

On some services, the flood is no longer theoretical. Deezer has publicly said it now receives 50,000+ fully AI-generated tracks per day, about 34% of its daily deliveries—and later raised that to 60,000+ per day, roughly 39% of daily intake. Those numbers are not from Spotify. They’re from a rival platform willing to publish what most companies prefer to keep vague. But the scale is clarifying: the upload pipe is no longer constrained by human labor.

Spotify, for its part, has offered a different kind of evidence—less about how many AI songs arrive, more about what it has to remove. The company told The Guardian it had taken down 75 million “spammy” tracks in a 12‑month period, a stark admission that mass production plus mass manipulation can overwhelm even the largest distributor of music on earth.

The question listeners keep asking—sometimes in good faith, sometimes as a provocation—is blunt: If platforms are being buried under tens of thousands of AI tracks a day, does that dilute what real artists earn? The answer is more precise, and more unsettling: the system is not “diluted” by volume alone. It’s distorted when volume is paired with gaming—strategies that manufacture plays, multiply payable events, or hijack attention through impersonation and metadata tricks.

Spotify has tried—often unsuccessfully—to correct one misunderstanding: there is no fixed per-stream rate. Spotify’s own “Loud & Clear” materials emphasize that payouts are a function of total revenue and total listening, varying by market, plan type, advertising rates, and the mix of subscribers and ad-supported users.

### The pro‑rata pool, in plain math

Under the prevailing pro‑rata model used by major digital service providers, the platform aggregates subscription and ad revenue, subtracts its costs/fees, and distributes the remainder to rightsholders based on share of total streams. Think of the system as a big monthly pie:

- Total eligible revenue becomes the royalty pool.
- Each track (really, each rightsholder’s catalog) receives a slice proportional to its share of total listening.

That structure leads to a crucial point: the number of tracks uploaded is not, by itself, what changes your payout. If nobody listens to the flood, it doesn’t matter—mathematically, a silent track has a share of zero.

### Where the “AI flood” becomes a royalty problem

The trouble begins when industrial-scale uploading is paired with industrial-scale behavior that produces listens—or the appearance of them. Spotify has publicly framed recent policy changes as targeting “gaming” tactics that generate “huge volumes of low-earning tracks” and drain money and attention from the system.

In other words, the risk is not “more songs,” but more payable events that are cheap to manufacture.

If pro‑rata math makes you think dilution is impossible without real listening, you’re already halfway to the uncomfortable truth: the main threat is not creative abundance. The main threat is fraudulent or engineered listening at scale.

Spotify’s September 2025 disclosure, reported by The Guardian, that it removed 75 million spammy tracks in 12 months gives the clearest signal of pressure on the system. Even without a daily AI-upload figure, that removal number suggests the platform is fighting not just low-quality music, but behavior patterns that violate rules or exploit them.

### Three ways payouts can be “gamed”

Spotify’s own framing of reforms points toward the most common mechanics:

- Fraudulent streaming: Plays manufactured by bots, clickfarms, or coordinated schemes that exist to trigger payouts rather than reflect genuine listening.
- Short-track and ambient/noise strategies: Structures designed to maximize “stream events” per minute, turning time into a higher count of payable actions.
- Impersonation and metadata manipulation: Look-alike artists, misleading titles, or confusing metadata that divert plays intended for legitimate artists.

None of these require convincing a listener to love a song. They require convincing the system to count something as a play.

### Why this matters to ordinary musicians

For working artists, the harm is not only financial. It’s also about discoverability and trust. A streaming catalog that can be cheaply polluted forces legitimate creators to compete against an infinite supply of content engineered to exploit recommendation systems and payout rules.

Platform cleanups help, but the scale implied by mass removals suggests a long, ongoing contest.

The headline-ready statistic—50,000 AI songs a day—has become a kind of shorthand for creative overproduction. But precision matters, and the cleanest public number comes from Deezer, not Spotify.

Digital Music News reported Deezer’s claim that it was receiving 50,000+ fully AI-generated tracks per day, about 34% of daily deliveries. A later Deezer press release dated January 29, 2026 raised the figure to 60,000+ per day, about 39% of intake, and referenced demonetization actions.

### What we can and cannot say about Spotify

What we can say from authoritative, cited reporting and Spotify’s own materials:

- Spotify has well over 100 million tracks on the service, according to Spotify for Artists.
- Spotify says it has removed 75 million spammy tracks in a year, per The Guardian (Sept. 25, 2025).
- Spotify has not, in the public materials cited here, published a daily number of AI uploads comparable to Deezer’s disclosure.

What we cannot responsibly say: that Spotify receives “50,000 AI songs per day.” Many internet retellings effectively paste Deezer’s figure onto Spotify because Spotify is the biggest target in people’s imaginations. That makes for an easy narrative—and sloppy journalism.

### Why Deezer’s numbers still matter to Spotify’s story

Deezer’s disclosure functions as a proxy for the industry’s direction of travel. If one major service sees tens of thousands of AI tracks daily, it’s reasonable to infer that other large services are also facing a surge. But inference isn’t evidence, and readers deserve the difference spelled out.

The more useful takeaway is structural: when creation is automated, enforcement and allocation rules become the real battleground.

On April 1, 2024, Spotify implemented a set of changes it described as “modernizing our royalty system.” The company’s argument was specific: money that had been scattered across vast numbers of low-engagement tracks—often exploited through gaming—would be reallocated to “emerging and professional artists.” Spotify said the shift would redirect “tens of millions” of dollars.

Spotify also emphasized something it wanted artists to hear clearly: these were allocation rules, not a reduction in the overall pool. The company’s stance is that it did not increase its own take; it changed who qualifies for slices of the pie.

### The minimum stream threshold (and what it signals)

Spotify’s public blog states that tracks below a minimum annual stream threshold would no longer generate recorded-music royalties, and that money previously distributed as tiny payments would be redirected to eligible tracks. Many summaries widely cite the threshold as 1,000 streams per year per track, and Spotify’s materials discuss the policy and its intent.

The policy isn’t just an accounting tweak. It’s a statement about what the platform considers economically meaningful activity—and an attempt to remove incentives to upload millions of tracks that earn pennies while creating administrative and fraud-monitoring burdens.

### Minimum track length and “noise” strategies

Spotify also described a minimum track length policy aimed at “noise recordings.” The logic is straightforward: if payouts are triggered by “streams,” then ultra-short files can be structured to produce more stream events per minute, especially in contexts like sleep playlists, ambient listening, or algorithmically generated background use.

Reducing that incentive is a form of rulemaking—less about punishing creators, more about preventing a particular exploit from being profitable.

Spotify’s reforms have a plausible public-interest case: less money wasted on microscopic payments and less incentive for spam. Still, skepticism is warranted, and not all criticism is bad faith.

### The “pro” argument: fewer incentives to game, more money for engaged music

Spotify’s stated rationale is that tiny payouts spread across millions of underperforming tracks create a perverse incentive structure. Removing monetization below a threshold aims to:

- discourage industrial-scale dumping of content designed to skim micro-royalties,
- reduce administrative complexity,
- concentrate payments on tracks with demonstrable engagement.

If the company’s claim is accurate—that these changes reallocate tens of millions—then the impact could be material for artists hovering above the minimum engagement bar.

### The “con” argument: thresholds can punish the smallest legitimate creators

A minimum threshold does not distinguish between spam and a genuine new artist with a tiny audience. A track can fail to reach 1,000 annual streams for ordinary reasons: niche genre, limited marketing, regional barriers, or simply being new.

Critics can reasonably ask whether the policy shifts the system toward a more professionalized tier—rewarding those who already have enough reach to clear a bar, while leaving hobbyists and micro-scenes with less.

Spotify’s implied counterargument is that the money below the threshold was already negligible at the track level and that the policy targets systemic abuse rather than bedroom artists. Both can be true: the dollar amounts per track might be tiny, and the symbolic message—your work doesn’t “count” until it performs—can still sting.

### A deeper issue these reforms don’t solve

Even perfect allocation rules don’t fix the fundamental vulnerability of pro‑rata systems: if fraud can manufacture a meaningful share of total listening, fraud can still siphon money. Thresholds and length rules may raise the cost of abuse, but they won’t eliminate it.

The most revealing details in this era are not philosophical. They are operational: what platforms are doing when the upload pipeline goes feral.

### Deezer: counting the flood, then cutting off the cash

Deezer’s public communications did two unusual things. First, they quantified the scale of AI intake—50,000+, later 60,000+, AI-generated tracks per day—turning a vague fear into a measurable flow. Second, they referenced demonetization, indicating a willingness to separate “available to listen” from “eligible to earn.”

That distinction may become the industry’s defining compromise. Platforms can host an enormous amount of content while restricting what can extract money from the royalty pool.

### Spotify: scale, then subtraction

Spotify’s data point—75 million spammy tracks removed in a year—tells a different story. Spotify’s catalog is vast (it says well over 100 million tracks), and moderation at that scale becomes an exercise in triage. Removal is a blunt instrument, but sometimes blunt tools are what the situation demands.

The platform’s April 2024 “modernizing” changes can be read as a complement to takedowns: instead of only playing whack‑a‑mole after the fact, Spotify is trying to make some kinds of abuse less profitable by design.

The AI era doesn’t only pose aesthetic questions. It forces a reckoning with measurement and incentives. For readers who want to understand where this goes, a few signposts matter more than the daily upload count.

### For artists: focus on engagement integrity, not volume

The new policies implicitly reward sustained listening and penalize strategies that look like mechanical extraction. Practical implications:

- If your track can’t reach minimum engagement thresholds, royalties may be negligible anyway—so treat streaming income as a lagging indicator, not the foundation.
- Metadata accuracy matters more than ever; impersonation and look‑alike tactics thrive in ambiguity.
- Sudden spikes from suspicious sources can become a liability if platforms interpret them as manipulation.

### For listeners: the catalog may grow, but trust is the scarce resource

The user experience risk is subtle: when spam rises, recommendation systems can become less reliable. Platforms can counteract that with detection and demotion, but the arms race is ongoing.

Listeners can help by treating playlists and recommendations with a slightly more discerning eye—following artists directly, saving albums, and supporting work that signals human intention and craft.

### For the industry: the fight is over rules, not rhetoric

The fundamental debate is moving from “AI is coming” to “what earns money.” Expect more of the following:

- Eligibility thresholds (streams, duration, or other engagement metrics),
- demonetization regimes for suspicious or low-value content,
- increased emphasis on fraud detection and enforcement.

These are governance choices. They will shape who can make a living in music as much as any model architecture will.

The fear that “AI songs dilute artist royalties” is a half-true headline attached to a more complicated system. In a pure pro‑rata model, adding songs doesn’t reduce payouts unless those songs attract listening. The real danger arrives when listening itself becomes cheap to manufacture.

Deezer’s disclosure—50,000+, later 60,000+, AI tracks per day—shows what happens when creation becomes automated. Spotify’s response—rule changes effective April 1, 2024, plus 75 million spammy tracks removed in a year—shows what happens when monetization and moderation collide at planetary scale.

The streaming economy is entering a phase where the central question isn’t “How much music can we host?” It’s “What kinds of activity deserve to be paid?” That question is ethical, economic, and political all at once. Platforms will answer it with thresholds, length rules, demonetization, and enforcement—and with every answer, they’ll redraw the map of who gets to be a professional artist.