The Royalty Heist
One man, 1,000 bots, and the quiet drain on every real artist's payout
You spent weeks making the mix sound honest. Every transient revealed, every frequency earning its space, every decay controlled with the kind of patience that comes from actually caring. Then you uploaded it. It went into the pool with everyone else’s work, waiting for streams to translate into something.
Meanwhile, a 54-year-old musician in North Carolina was collecting from the same pool with songs that never existed and listeners who were never real.
On March 21st, Michael Smith pleaded guilty to defrauding streaming platforms out of more than $10 million. The mechanism: AI-generated tracks, uploaded by the thousands. Bot accounts, over a thousand of them at peak, streaming those tracks billions of times across Spotify, Apple Music, Amazon Music, and YouTube Music. Seven years. Every stream a fraction of a cent pulled from the shared pool your real music also draws from.
How Royalty Pools Actually Work
The streaming royalty isn’t a flat rate. There’s no “you get $0.004 per stream” rule that’s fixed regardless of what else is happening on the platform. It’s a pool.
Here’s the structure: every month, platforms take their total royalty budget and divide it proportionally among all streams. More streams on other tracks means a smaller percentage for yours. The pie is roughly fixed. The number of slices keeps growing.
When Smith’s bots added 661,440 artificial streams per day to that pool, the denominator for every real artist’s calculation got larger. Your stream percentage went down. Not dramatically on any given day. But consistently, for seven years, across four platforms.
Smith emailed himself the math in 2017. He estimated each bot account could stream 636 songs per day. With 1,040 accounts: 661,440 daily streams. Average royalty rate of half a cent per stream. Annual earnings: over a million dollars. His own projection.
In a 2018 email to coconspirators, he wrote: “to not raise any issues with the powers that be we need a TON of content with small amounts of Streams.” Then: “We need to get a TON of songs fast.”
He was optimizing for detection avoidance. That’s what took seven years.
The Scale Problem Is Bigger Than Smith
Smith’s case is vivid because it has a face, a name, and a guilty plea. But he’s not the scale problem. He’s a symptom of it.
In January 2026, Deezer reported receiving more than 60,000 fully AI-generated tracks every day. Not weeks, not months. Every day. The IFPI’s 2026 Global Music Report followed with the number that explains why this matters for royalties: 85% of streams on AI-generated music across Deezer in 2025 were fraudulent: bots, not listeners. Up from 70% in mid-2025.
The mechanism scales without effort. The same infrastructure that let one person extract $10 million can be replicated by anyone with a credit card and working knowledge of cloud services. The barrier isn’t technical. It’s the willingness to do it.
Every fraudulent stream that hits the pool is a fraction of a cent redistributed from your release to a file that’s never played in a room, never felt by anyone, never meant anything to its creator.
What’s Being Built in Response
Two things are shifting, and they’re worth tracking.
Spotify announced Artist Profile Protection this month: a system that lets artists review and approve what music appears attributed to their profile. It’s aimed at a specific fraud vector: AI-generated or stolen tracks uploaded under a real artist’s name. The framing from Spotify leadership: “protecting artist identity is a top priority for 2026.” That’s a policy statement, not just a feature launch.
The second signal: the IFPI’s 2026 report documents major labels forming the largest AI licensing coalition in the industry’s history. The goal is to establish terms for how AI systems can legally use catalog music, which implicitly creates pressure on platforms to distinguish licensed AI content from unlicensed bot operations.
Neither of these solves the problem. They’re early infrastructure. But they represent something that wasn’t true two years ago: platforms and rights-holders are starting to build walls. The pressure on distributors to verify human origin is increasing. The cost of running a bot farm is going to rise as detection improves.
The Smith plea matters partly because it establishes precedent. Federal conviction for streaming fraud. $8 million in forfeiture. Five years maximum. That’s a signal to others calculating the risk-reward of Smith’s playbook.
The Part Most Analysts Miss
Here’s the second-order effect that doesn’t make the trade press: if platforms are being forced to differentiate human music from bot-generated content, that differentiation creates value for the human side.
Discovery algorithms that can identify authentic listening patterns will surface music to actual listeners with preferences, tastes, and memories attached to songs. The AI slop floods the general pool, but the verified human content goes where real ears go.
You’ve spent years developing an ear for what makes a mix feel alive: the transient that hits right, the saturation level that reveals presence without smearing it, the decay that disappears at the exact right moment. That care doesn’t show up in a metadata field. But it shows up in listening behavior. Listeners who respond to authentic production complete songs, save tracks, return to albums, follow artists.
That behavior is what algorithms are being trained to distinguish from bot-pattern streams. The same cleanup that removes Smith’s fraud also improves the signal quality of the discovery systems your music runs through.
The royalty pool problem is real and ongoing. But the structural response is creating a landscape where being human, making music that sounds like a person made it, is becoming a competitive advantage rather than a baseline assumption.
One Number to Carry
Smith’s math was clean: 661,440 fake streams per day. By his own calculation, that was $3,307 per day. Over seven years, prosecutors say it added up to more than $10 million, extracted from a pool built on the assumption that streams represent listeners.
The assumption was always wrong. The pool was always vulnerable. What’s changing is that the platforms know it, the labels know it, and the DOJ has now drawn the criminal line.
Make the music sound honest. It’s the only thing that holds its value when everything else gets cleaned up.
Resources & Further Reading
On the Michael Smith case:
Musician admits to $10M streaming royalty fraud using AI bots — BleepingComputer (March 2026). Detailed coverage of the Smith plea, including the court documents, financial projections, and the email chains that show how the operation was designed to evade detection.
US man pleads guilty to defrauding music streamers out of millions using AI — The Guardian (March 21, 2026). Clean overview of the plea and the precedent it sets as the first federal conviction for AI-assisted streaming fraud.
On the scale of the problem:
Deezer says up to 85% of its AI-music streams are now fraudulent — Music Ally (January 29, 2026). The primary source for the 85% fraud rate and the 60,000-tracks-per-day figure, with month-by-month context on how fast the numbers climbed through 2025.
IFPI Report 2026: Global Music Revenues Surpass $30 Billion Mark — Billboard (March 2026). The industry-level view: revenue growth, AI fraud data, and the labels coalition forming in response.
On platform response:
Spotify Strengthens AI Protections for Artists, Songwriters, and Producers — Spotify Newsroom (September 2025). Spotify’s official policy announcement, including tighter upload filters and anti-impersonation measures.
Spotify takes its first major step in tackling AI slop — TechRadar (March 2026). The Artist Profile Protection feature: what it does, what it doesn’t do, and why Spotify is calling this a “top priority for 2026.”