A Spotify executive sat for an interview about AI music, catalog fraud and royalty payouts, and gave the kind of answers that are only useful if you read them carefully. He never lied. He just answered the questions he was asked, and the questions he was asked were about fraud — not about the machinery underneath it.
The numbers he offered are real and worth having. Over the past twelve months, he said, Spotify removed roughly 75 million “spammy” tracks. He defined the term on the spot, and the definition matters: a spammy track is one “essentially trying to deceive Spotify or trying to game the system” — regardless of what technology was used to make it. That is not the same thing as an AI track. He was clear about the gap, and it is a fair distinction.
His other headline figure: less than 1% of streams on the platform are deemed to be artificial. He added that this is an industry-wide fight, not a Spotify one, and that the company had spent the year rolling out a new spam filter, tightened impersonation policy, and an industry standard for AI credits so a listener can see what they are hearing.
His framing, in his own words
“We don’t tell artists how to create.” That was the position, repeated more than once: AI is a tool, and Spotify’s job is to facilitate the good uses and defend against the bad ones. Pressed on where the line is, he drew it at volume and deception — someone delivering tens or hundreds of tracks a day to the platform, or using boosting mechanisms to move chart or vanity numbers. That is a behaviour test, not a technology test, and it is the most defensible thing he said all interview.
On whether AI music is actually taking the audience, his answer was blunt: prompt-generated music shows “very, very low engagement.” Artists built on it spike and then nobody listens. The share of listening on the platform that comes from AI-generated tracks, he said, is “super, super, super low.”
The word that gave it away
Later the conversation turned to derivatives — fan-made remixes and altered versions, the UGC layer that AI makes cheap. His position was that any tools Spotify builds for that would arrive in partnership with rights holders and on an opt-in basis. Then the host said the quiet part, and it is the most revealing exchange in the hour:
“Opt-in doesn’t scale very well.”
The executive did not disagree. Asked whether that means every individual artist signed to Sony, Warner and Universal gets their own opt-in choice, he said that is where the conversation is, and that getting there is “a logistical nightmare.” Read those two answers together. A right that cannot be administered individually is not, in practice, a right held by an individual.
The one he agreed with that he didn’t have to
Toward the end the host made a structural point: whatever gets built here, the master rightsholder takes a much larger share of the pie than the songwriter community, and that has been a problem for years. The executive agreed with him.
That is not a slip. It is the industry’s own baseline, and it is the reason the AI question is a *songwriting* question more than a recording one. AI is very good at the thing the songwriter is paid least for. The composition — melody, lyric, structure — is the layer with the thinnest claim on the money and the cheapest thing to generate.
He brought his own test, and it convicts
The best moment is one he volunteered. Talking about speed-altered “sped up” versions that took over the platform a few years back, he asked three questions about them: “Did someone ask the artist’s permission? Was the artist compensated? Did it infringe on the artist’s rights?”
That is the right test and it applies to everything. It also applies to content that no one has to break into your catalogue to make — the library track, the pseudonymous instrumental, the anonymous mood piece that never earned a royalty because it never had a rightsholder sitting on the other end of the wire. Nobody infringed anything. There was nobody to infringe on.
What the interview didn’t cover
Three things sit outside the conversation and frame everything inside it.
First, the profit. Spotify reported its first full year of profitability in 2024 — net income of roughly €1.14 billion — achieved alongside price increases, subscriber growth and the elimination of around 1,500 jobs. AI was not the cause of that. It has never needed to be.
Second, the floor. Since April 1, 2024, a track needs 1,000 streams in twelve months to generate recording royalties at all. Spotify’s own line is that this does not change the royalty pool. What it changes is who is standing in the room — and the bottom of the catalogue is precisely where the next generation of artists, and the cheapest content, both live.
Third, Perfect Fit Content. Liz Pelly’s January 2025 investigation in Harper’s documented an internal Spotify program filling mood playlists with generic library music made by producers under pseudonyms for a flat fee — tracks that generate plays without generating a royalty negotiation. That is the same problem as AI, solved without AI, years earlier.
The switch, in Spotify’s own writing
There is a document that makes the point better than any of this, and it arrives in artists’ inboxes. It is Spotify for Artists’ own email marking ten years of Fresh Finds, the editorial playlist program. By the company’s own count it has featured nearly 70,000 emerging artists from well over 100 countries, and the promotional claim is the interesting part:
“On average, one month after being added to Fresh Finds for the first time, artists see their monthly listeners more than double.”
Read that as a gift and it is a genuine one — tens of thousands of artists got their first real audience from it. Read it as a business document and it is an admission. The platform is stating, in writing, that it can double an artist’s audience in thirty days with a decision made in an editorial meeting, and that the artists it selects are chosen by “a team of Spotify editors who review artist playlist pitches and leverage data insights.”
That is the switch. Not AI, not fraud, not the 1%. A platform that can double your listeners in a month with one decision can also not do that, and the difference between those two outcomes is not a matter of how good the song is. It is a matter of who is holding the lever. Everything else in the AI conversation — the credits, the spam filter, the opt-in — is a policy built around a thing that was already true.
SPLASH’S READ
SPLASH'S READ: The executive’s statistics are true and they answer a question nobody’s career depends on. Less than 1% of streams being fake is a statement about fraud. It is not a statement about payout. A platform can be genuinely aggressive about fraud and still be the place where the cheapest possible audio is the most profitable audio, because the royalty on a track with no rightsholder is zero and the royalty on yours is not.
That is why the “is AI music real music” argument goes nowhere. Music was never priced by quality. It was priced by who can claim it. The whole history he points at makes the case better than he meant it to: the UK Musicians’ Union tried to ban synthesizers in the early eighties, and the machines won not because they sounded better but because someone could own the output more cheaply. He said it himself in the first minute: Spotify was born out of piracy and does not want to go back there. Fair. But the industry it replaced was not killed by people stealing music. It was killed by a business model that made music nearly free to distribute and then paid the people who made it out of whatever was left.
So the number to watch is not 1%. It is the threshold, the playlist, and the split — three switches the platform operates and the artist does not. A right that arrives opt-in and cannot be administered one artist at a time is not a right. It is a press release.
Watch: Top Music Attorney.
Sources & claims: The Manager’s Playbook,
Spotify Exec Breaks Down AI Music, Fraud & Royalties, published May 14, 2026 — an interview with a Spotify executive who is
not named in the upload, so every quotation here is attributed to “the executive” rather than to a person.
Disclosure: the episode contains a mid-roll advertisement for DistroKid, a music distributor, which is relevant context for a conversation about distribution and royalties. Figures taken from the interview and attributed to the executive:
75 million spammy tracks removed in twelve months;
less than 1% of streams deemed artificial; AI-generated tracks a
“super, super, super low” share of listening; the definition of a spammy track as one attempting to deceive or game the system.
Not Spotify’s figure: the host’s aside that
60,000 new AI records are released daily — he attributes it to “I saw somewhere” and it is not verified here. Independent context added and sourced separately: Spotify’s
Q4 2024 earnings (February 4, 2025) for the
first full year of profitability, €1.14 billion net income, and contemporaneous reporting (Fortune, Variety) for the
~1,500 job cuts;
Spotify for Artists / Music Business Worldwide (April 1, 2024) for the
1,000-stream royalty threshold;
Liz Pelly, “The Ghosts in the Machine,” Harper’s, January 2025 for the Perfect Fit Content program. The synthesizer/Musicians’ Union episode and the “born out of piracy” framing are the executive’s own remarks, not independent findings.
No allegation of illegal conduct is made against Spotify here — the argument is about incentive structure, not criminality. The
Fresh Finds figures — ten years, nearly 70,000 artists, 100+ countries, monthly listeners more than doubling within one month, and the ‘team of Spotify editors’ line — are quoted from a
Spotify for Artists promotional email received November 2026 marking the program’s tenth anniversary. It is the company describing its own curation; it is not an independent audit.