We are the music makers

News · industry — By WATMM Desk —

Sync in the AI era: why a clean chain of title is becoming the premium product

AI tracks are flooding the cheap end of production music, but a music supervisor can't clear or insure a cue whose authorship and training data nobody can document. Here is how sync money works, and why well-documented human catalogues are gaining value.

Key facts

  • What happened: Deezer said on 21 July 2026 that fully AI-generated tracks passed 50% of its new uploads at their June peak, nearly 90,000 a day, while drawing only 1-3% of streams.
  • What is at stake: the US Copyright Office says prompts alone do not make someone an author, and ASCAP, BMI and SOCAN will not register fully AI-generated works, so a pure AI cue can have no clear owner and no performance back end.
  • Why it matters: a sync licence is only as good as the chain of title behind it. As the cheap library tier fills with AI tracks, documented, human-made catalogues with clean splits become the scarce thing.

Sync, the licensing of music into film, TV, ads, games and online video, has always been a paperwork business that happens to involve songs. Generative AI is testing that. Deezer said in July 2026 that fully AI-generated tracks made up more than half of its new uploads at their June peak, nearly 90,000 a day. In November 2025 the figure was over 50,000 a day and about 34% of deliveries. A lot of that music is exactly the kind of mood-and-tempo bed that the cheap end of sync runs on.

The argument of this piece is simple. AI can make the audio cheaply. It can't make the paperwork, and in sync the paperwork is what a buyer pays for.

How sync money works

Two rights, two licences. Any recording has two copyrights: the composition (melody, lyrics, harmony), usually controlled by a songwriter or publisher, and the master (the specific recording), usually controlled by a label, an independent artist or a production music company. Putting a track in a video needs both a sync licence for the composition and a master use licence for the recording, as Artyfile's explainer sets out. If either side can't be cleared, the track can't be used.

Agents and libraries. There are two broad routes in. Sync agents pitch music to supervisors, ad agencies and producers and negotiate each deal. Tools 4 Music puts their commission at 25% to 50%, with a 50/50 split on the sync and master fees the common standard. Libraries hold pre-cleared catalogues that buyers search and license directly. Non-exclusive libraries (the same track can sit in many) typically pay the artist 35% to 50%. Exclusive libraries typically pay 40% to 60% and push the tracks harder. The same guide says no reputable library should ask for copyright ownership.

The upfront fee. Vendor estimates vary widely and should be read as rough guides. Tools 4 Music suggests $2,000 to $20,000 for a TV sync placement and $50 to $2,000 for a library placement. Artyfile quotes $500 to $5,000 for an original by an independent artist, and $10,000 to over $500,000 to clear a well-known song.

The back end. Broadcasters and streaming services pay performing rights organisations (PROs) a blanket fee. Productions file cue sheets listing every piece of music used, and cue sheets are how PROs track which writers and publishers to pay. For a library composer, that means a cue can keep earning performance royalties every time it airs, long after the upfront fee is spent. But that only works if the cue is registered to an identifiable writer.

Where AI breaks the chain

Each of those steps assumes somebody can say who made the music and who owns it. A fully AI-generated track struggles on every count.

Authorship. The US Copyright Office's January 2025 report concluded that protection can't rest on "the mere provision of prompts". AI-assisted work can still be protected where a human determined enough of the expressive elements. Suno's own help centre says that Suno owns songs made on its free Basic plan, that Pro and Premier subscribers own theirs, and also that music made 100% with AI would not qualify for US copyright protection, because writing the prompt does not count as creating the song. So a buyer can be "licensed" a track whose copyright may not exist.

The back end. On 28 October 2025, ASCAP, BMI and SOCAN aligned their policies. They now register works that combine AI-generated material with human authorship, but fully AI-generated works can't be registered with any of the three. A pure AI cue has nowhere to go on a cue sheet in North America.

Insurance and indemnity. This is the part supervisors feel most. Chris Castle, who writes Music Technology Policy, argues that a synthetic track can be a "rights black box". The people with a claim on it may be unknowable, so a supervisor can't investigate, clear or insure against them. In his account, errors-and-omissions (E&O) underwriters now ask AI-specific clearance questions that, in nearly all cases, make cover very hard to get. Because E&O cover is often part of what a producer must deliver, failing to get it can hold up a licence fee or minimum-guarantee payment. His checklist asks whether the licensor gives a warranty and an indemnity covering AI elements and training data, and whether it can afford to honour them. That is his analysis rather than an industry survey, but it matches how clearance works: the risk sits with whoever signs.

The trade press sees the same split. The Sync Brief argues AI is winning at the bottom of the market, in high-volume background music for YouTube, social, corporate video and low-budget work, while prestige TV, film and national advertising stay human-led. It says it knows of no AI-generated music in those top placements "at least yet".

What we think

The useful way to read this is that sync is splitting into two products. The bottom tier sells audio, and AI will win it on price. The top tier sells certainty: a named writer, signed splits, a registered work, a master owner who will warrant it and an insurer who will cover it. AI doesn't compete in that tier, because certainty is exactly what an AI track without documentation can't supply.

Some numbers make the gap concrete. At Deezer's "nearly 90,000 a day", that is roughly 2.7 million fully AI tracks a month arriving at one service. That is our arithmetic, and only some of those tracks will ever be pitched for sync. Deezer says those tracks draw only 1-3% of streams, and that up to 85% of streams on them in 2025 were fraudulent. Volume isn't demand. For a supervisor, a library that adds thousands of anonymous tracks a week makes the search slower and the clearance riskier.

The back end matters more than it looks. A library cue that is fully AI-generated can't be registered with ASCAP, BMI or SOCAN, so it earns no writer performance royalty from them. Its only income is the upfront fee, which is already being pushed towards zero. Human library composers earn on both. Our view: the AI flood squeezes upfront library fees, but it raises the relative value of the long-tail royalty that only documented human work can collect.

There is a catch for artists. When buyers want certainty, they ask for warranties that a track is human-made and uses no uncleared material, backed by an indemnity. Big publishers can carry that risk. A solo composer signing a library's standard terms may be taking on a liability far bigger than a $200 placement fee. Read the warranty and indemnity clauses before the fee, and push back on terms that make you insure the library's whole catalogue. Libraries that knowingly mix undocumented AI tracks into the same catalogue as human ones, then ask human composers for blanket warranties, are pushing their own risk downhill.

What "clean chain of title" means in practice, for a working composer:

  • Signed split sheets for every co-write, agreed before the track is pitched.
  • Registration with your PRO (and, in the US, The MLC), and an ISRC on every master, as The Sync Brief also recommends.
  • Session files, stems and dated drafts kept as evidence of human authorship. We think these will become the provenance record supervisors ask for.
  • A written note of any AI tools used and on what. Partially AI works can still be registered, but only if you can show where the human authorship is.

Licensed, closed AI models are the obvious answer for the AI side, and some exist, such as the walled-garden route Udio took. Castle's workable path is similar: a model trained only on owned or licensed material, with provenance tracking and an indemnity behind it. Until such models can hand a supervisor the same paper trail a human catalogue can, the premium in sync will go to whoever can prove who made the music.

Sources

Sources

  1. Deezer: AI music has surpassed 50% of new music uploads for the first time, Deezer Newsroom (primary source). Accessed 23 Oct 2026.
  2. Deezer/Ipsos survey on AI music, Deezer Newsroom (primary source). Accessed 23 Oct 2026.
  3. Copyright Office Releases Part 2 of Artificial Intelligence Report, US Copyright Office (primary source). Accessed 23 Oct 2026.
  4. ASCAP, BMI and SOCAN Announce Alignment on AI Registration Policies, SOCAN (primary source). Accessed 23 Oct 2026.
  5. Do I have the copyrights to songs I made?, Suno Help Center (primary source). Accessed 23 Oct 2026.
  6. Why AI tracks put music supervisors in an impossible position, Music Technology Policy (Chris Castle). Accessed 23 Oct 2026.
  7. What AI music means for your next licensing brief, The Sync Brief. Accessed 23 Oct 2026.
  8. Sync licensing companies vs. music libraries, Tools 4 Music. Accessed 23 Oct 2026.
  9. What is a sync license? Cost and how to get one (2026), Artyfile. Accessed 23 Oct 2026.
  10. What is a cue sheet?, Soundstripe. Accessed 23 Oct 2026.

Researched with AI assistance, written and checked by WATMM Desk.

Tags

industrysynclicensingaimusic-supervisorscopyright

Subscribe to We are the music makers

One email a week: the music-tech and AI-in-music news that mattered, and what we think about it.

Sign-ups open soon.