Two things are true about AI and music industry right now, and they pull in opposite directions. The supply of AI generated tracks is exploding. What listeners say they want is not. Somewhere between those two facts, labels, platforms, and licensing teams are trying to build a working framework in real time, and nobody has finished the job.
This is not a hype piece and it is not a panic piece. It is a look at what has actually happened so far in 2026, drawn from platform disclosures, trade body reports, and the settlement filings that are quietly rewriting how music gets licensed. Then it gets specific about what that shift means for the people who manage sync deals and licensing pipelines day to day.
THE SUPPLY SIDE MOVED FASTER THAN ANYONE’S FRAMEWORK
Start with the number that changed the conversation. Deezer, the only major platform that has published a breakdown, reported that AI generated tracks accounted for 44% of everything uploaded to the platform daily by April 2026, roughly 75,000 tracks a day. In January 2025, that figure was 10,000 a day.
Those AI uploads represent 44% of daily volume but only 1 to 3% of actual Deezer streams, and Deezer reported that up to 85% of the streams those tracks did generate in 2025 were flagged as fraudulent, driven by stream farming bots rather than real listeners.
Spotify, Apple Music, and Amazon Music have not published comparable figures, which is itself worth noting. What Spotify has disclosed is the cleanup side: it removed more than 75 million spammy tracks in the twelve months ending September 2025 and adopted the DDEX standard for AI music credit disclosure.
The more telling number may be this one: a LANDR survey of over 1,200 artists conducted in the second half of 2025 found that 87% already use AI somewhere in their creative workflow, whether that is mastering, stem separation, or sample generation. AI in music is not an outsider technology waiting to arrive. It is already embedded in how a large share of working musicians produce their tracks.
THE METADATA LAYER IS WHERE SYNC TEAMS FEEL IT FIRST
The training data lawsuits get the headlines, but a quieter AI shift is changing day to day work inside production libraries and sync agencies: automated tagging. Tools built for this specific task, including Cyanite and Music.AI, listen to a track and generate genre, mood, tempo, and instrumentation tags automatically, replacing a process that used to be manual, slow, and inconsistent enough that the same track could pick up three different mood tags from three different taggers.
That matters more, not less, as AI generated supply floods in. A brief that used to take an afternoon to fill from a catalog of a few thousand tracks now has to be filled from a catalog where new material, AI assisted or not, is arriving faster than any human team can listen to and tag by hand.
SPEED WITHOUT A HUMAN EAR IS STILL A BOTTLENECK
Vendors serving this space, including TagTeam Analysis, still route AI generated tags through a human listener before they go live, treating automation as a speed layer rather than a replacement for a trained ear on genre sensitive attributes like mood and usage.
For a sync or licensing team, metadata and deal terms solve two different problems. Metadata determines whether the right track gets found for a brief in the first place. What happens after it is found, consent status, usage terms, who has agreed to what, is a separate layer that a tagging tool was never built to hold. Both layers need to work together for a sync process to keep pace with the current volume, which is exactly why licensing teams are having to think about pipeline visibility and catalog discovery as two connected problems rather than one.
THE LEGAL FRAMEWORK IS BEING WRITTEN DEAL BY DEAL
On June 24, 2024, the RIAA filed parallel suits on behalf of Universal Music Group, Sony Music, and Warner Records against Suno and Udio, seeking statutory damages of up to $150,000 per work for training on copyrighted recordings without authorisation. Sixteen months later, the settlements started arriving, and they are not identical.
- UMG settled with Udio on October 29, 2025: a licensing deal, opt-in artist compensation, and a joint platform launching in 2026 that keeps generated tracks inside the platform rather than allowing downloads.
- Warner settled with Suno in November 2025, described as a first of its kind licensing partnership, with Suno agreeing to retire its current models in favor of licensed versions.
- Warner also settled with Udio on November 19, 2025, and unlike the UMG deal, Warner’s agreement allows downloads.
- BMG signed a global strategic alliance with Suno on August 12, 2026, covering both its recorded and publishing repertoire. The deal settles Suno’s prior use of BMG’s catalog and requires artists and songwriters to opt in before their work is licensed to future models, making BMG the first major rightsholder to strike a deal with Suno since Warner’s agreement in November 2025.
- UMG and Sony remain in active litigation with Suno. Reporting in early April 2026 described talks as being at a hard impasse over that same download question.
A new front opened in January 2026, when a group of music publishers filed suit against Anthropic, alleging BitTorrent based piracy of more than 20,000 compositions with damages potentially exceeding $3 billion. It is a separate claim from the training data question and a sign that the legal exposure around AI and music is still widening, not narrowing.
Read across every settled deal so far and three structural features repeat: opt-in licensing rather than blanket catalog grants, per-use or per-stream compensation rather than flat fees, and in at least one case, walled garden distribution that restricts where a generated track can live. That combination, consent given per work rather than assumed, and payment tied to actual use rather than a lump sum, is the shape AI licensing is taking across the industry.
LISTENERS CAN’T TELL THE DIFFERENCE, AND SAY THEY WANT TO KNOW ANYWAY
A Deezer and Ipsos blind test of 9,000 respondents across eight countries found that 97% could not reliably distinguish an AI generated song from a human made one. At the same time, a Bain & Company poll conducted with The Hollywood Reporter, surveying roughly 2,000 US respondents, found that 62% said they were less likely to engage with AI generated music.
THE PARADOX IN ONE LINE
People can’t tell the difference in a blind test, but they say they don’t want it anyway. The same Deezer and Ipsos survey found that 80% of respondents want fully AI generated tracks clearly labeled.
That stated preference, not the blind test result, is what platforms are building policy around. Spotify’s September 2025 disclosure rules and YouTube’s July 2025 demonetization policy for mass produced content both respond to what people say they want, not what they can actually detect when they are listening.
WHAT THIS MEANS FOR WHO GETS PAID
The IFPI Global Music Report 2026 puts global recorded music revenue at $31.7 billion in 2025, up 6.4% year over year, with streaming now accounting for 69.6% of that total. That is the pool AI is being inserted into, and it keeps growing, which raises the stakes on who captures the AI generated share of it.
The CISAC and PMP Strategy global economic study projects that music creators could lose 24% of their revenue to generative AI by 2028, a cumulative loss of roughly 10 billion euros between 2023 and 2028, while the AI music output market itself is projected to reach 16 billion euros annually over the same period. In aggregate, the new market roughly offsets the losses. The distribution is the problem: that new revenue is flowing toward model providers, platforms, and infrastructure, not toward the songwriters and performers whose work trained the models in the first place. The opt-in licensing settlements described above are, in effect, the industry’s first attempt to correct that imbalance one deal at a time.
WHAT IT ACTUALLY CHANGES FOR A LICENSING OR SYNC TEAM
None of the above is abstract for the people who manage sync and licensing pipelines. It changes what a single track’s licensing status actually contains.
A track used to carry one licensing question: cleared or not cleared. It now carries several, and they can differ from one rights holder to the next: has AI training consent been given or withheld, is the track subject to a walled garden distribution restriction, does it require an AI disclosure label under a platform’s policy, and what per-use or per-stream terms were negotiated rather than a flat fee. Multiply that across a roster or a catalog of relationships and the operational load is not a catalog problem. It is a pipeline problem, and specifically a visibility problem: knowing at a glance where each deal, each rights holder, and each consent status actually stands, before a gap between them becomes a liability.
This is the exact kind of complexity a CRM built specifically for the sync and licensing workflow is meant to hold, as opposed to a generic sales tool retrofitted for the industry. Whether a team runs on SM Core, SM RockStar, or SMOH, the underlying need is the same: a system that tracks deal stage, relationship history, and negotiated terms per rights holder, so that when the next AI related clause lands in a contract, it has somewhere structured to live rather than becoming a note in someone’s inbox.
Curious what deal tracking built for this level of granularity actually looks like? Book a discovery call with the SMR team.
FREQUENTLY ASKED QUESTIONS
Is AI generated music now the norm on streaming platforms?
Not by consumption. Deezer is the only major platform to disclose figures, and while AI tracks made up 44% of daily uploads by April 2026, they accounted for only 1 to 3% of actual streams. Spotify, Apple Music, and Amazon Music have not published comparable numbers.
Have major labels settled with AI music generators?
Some have. Universal Music Group and Warner Music Group have both settled with Udio, and Warner has also settled with Suno. Universal Music Group and Sony Music remain in active litigation with Suno, with talks described as being at a hard impasse as of April 2026. BMG has also signed a global licensing alliance with Suno, announced August 12, 2026, covering both recorded and publishing rights.
Do listeners actually want AI generated music?
Stated preference and blind test behavior point in opposite directions. A Deezer and Ipsos blind test found 97% of listeners could not tell AI generated songs from human made ones, while a separate Bain & Company poll found 62% of US respondents said they were less likely to engage with AI generated music.
What does this mean operationally for a licensing team?
More deal types, more consent statuses to track per track, and more per-use terms that vary by rights holder. That raises the bar for pipeline visibility well beyond what a flat cleared or not cleared status used to require.
- IFPI Global Music Report 2026
- Deezer Newsroom, AI generated tracks represent 44% of new uploaded music
- Deezer and Ipsos blind test survey on AI music
- Bain & Company / Hollywood Reporter consumer attitudes survey
- LANDR survey on AI use in artist workflows
- Cyanite, AI Auto-Tagging for Music Catalogs
- TagTeam Analysis, AI-Assisted Music Tagging With Human Quality Control
- RIAA, Record Companies Bring Landmark Cases Against Suno and Udio
- Hollywood Reporter, Universal Music Group Announces Settlement With Udio
- Music Business Worldwide, Warner Music Group Settles With Suno
- Digital Music News, Suno and Universal Music settlement talks at an impasse
- IPWatchdog, Music Publishers File New Piracy Suit Against Anthropic
- CISAC, Global Economic Study on Human Creators at Risk From Generative AI
- Music Business Worldwide, Generative AI Output Market Forecast
- BMG and Suno Announce Global Strategic Alliance
- Music Business Worldwide, Suno inks global licensing deal with BMG
