ENTERTECH ANALYSIS | MUSIC STREAMING
‘Made With AI’ labels arrive this year as more than a third of monthly uploads are AI-generated and listening stays under 0.5%
Apple Music will label songs made with AI. The company emailed music industry partners on August 20 to say a “Made With AI” label will appear on the service later this year. It did not give a launch date.
In the email, Apple Music said content providers will be required to include AI Transparency Tags wherever AI was used to create a material portion of the content, including tracks that are AI platform generated, and defined AI-platform-generated content as anything primarily derived from a generative AI service. The Hollywood Reporter obtained the email and reported it first on August 20; Billboard and Variety reviewed the same document that day.
Apple Music did not say how the requirement will be enforced. The email said content providers will be important partners as the labels roll out, and that they are best positioned to know how their content was created and which tags apply.
Tags introduced in March become mandatory in five months… four elements covered
Apple Music unveiled AI Transparency Tags in an industry newsletter on March 4. The framework covers four creative elements — Artwork, Track, Composition and Music Video — with the artwork tag applied at album level and the track tag at track level. At launch the tags were optional, and Apple deferred to content providers on what counts as AI-generated. Where no tag was submitted, Apple treated the release as not AI-generated.
Two things change now. Submitting a tag becomes a requirement, and the information moves from Apple’s content delivery system to a label listeners can see. Music Business Worldwide noted in March that Apple’s framework depends on declaration at the point of delivery.
A third of uploads are AI-generated… listening stays below 0.5%
Oliver Schusser, who runs Apple Music, told Billboard that more than a third of monthly uploads are AI-generated songs, while actual listening to AI music remained under 1% — under 0.5% in the version of the remark cited by MacRumors. Schusser said Apple has an in-house detection system but still relies on content providers to disclose AI use.
AI share of uploads versus listening on Apple Music. Source: Oliver Schusser, Billboard interview; cited by MacRumors
The gap between upload share and listening share matches figures from other platforms. Deezer (디저) said in January it was receiving more than 60,000 fully AI-generated tracks a day, up from 10,000 when its detection tool launched in January 2025, then 30,000 in September and 50,000 in November. Deezer said it had detected and tagged 13.4 million AI-generated tracks in total; AI accounted for 44% of daily uploads in April and passed half of daily uploads by the summer. Deezer attributes the volume primarily to streaming fraud rather than creative intent.
Fully AI-generated tracks delivered to Deezer per day. Source: Deezer
Apple Music doubled its penalty fees for streaming fraud in January. The Hollywood Reporter reported the change first, and Schusser said the spread of AI content was a factor in the decision.
RIAA and IFPI proposed two labels in July… modeled on the explicit tag
Apple’s announcement came six weeks after a labeling initiative led by the Recording Industry Association of America (RIAA) and the International Federation of the Phonographic Industry (IFPI). On July 10 the groups introduced a program for marking AI music on streaming services in a manner similar to the explicit-content tag.
The proposal has two labels: “AI-Generated” for a recording created wholly or mostly by AI, and “AI-Assisted” for a recording that is substantially human-created but uses AI for some expressive elements. IFPI CEO Vikki Oakley and RIAA chairman and CEO Mitch Glazier said in a joint statement that fans want to know whether and how generative AI has been used in the music they listen to. The Recording Academy, SAG-AFTRA, A2IM, WIN, IMPALA and the Human Artistry Campaign signed on. The Wall Street Journal reported the initiative first on July 10.
Unit of disclosure: Apple Music tags versus the RIAA/IFPI labels
Based on Apple Music’s March newsletter and the RIAA/IFPI announcement of July 10. Source: Apple Music, RIAA, IFPI
Spotify labels the profile, not the song… ‘AI Persona’ plus removal from recommendations
Spotify said on August 11 that it will attach an “AI Persona” label to artist identities that are AI-generated. The badges appear from mid-September on profile banners and About sections, in search results, and on track rows in playlists. Spotify’s label attaches to the profile behind the music rather than to the music itself.
Spotify said that by default it will not include AI Personas in any editorial or algorithmic recommendations, and that listeners will not see their music in personalized recommendations unless they follow those profiles. Self-disclosure opened on August 11 through Spotify for Artists, and the company said it will not rely on self-disclosure alone: it will review profiles whose name and imagery appear to represent photorealistic AI-generated identities, starting with those above defined audience thresholds. Human artists who use AI in their process are not covered by the label.
Spotify introduced a separate track-level AI disclosure feature in April. Disclosure on individual songs is voluntary, and the company said it has seen tens of thousands of credits submitted daily.
AI music disclosure policies compared (as of August 2026)
Compiled from company announcements and reporting. Source: Apple Music, Spotify, Deezer
A label is disclosure; removal from recommendations is a sanction
Disclosure policy splits into two layers. One tells listeners how a track was made; the other reduces how often it is surfaced. Apple Music announced the first. Spotify’s exclusion from recommendations, effective from September, is the second. Given Deezer’s finding that much AI upload volume is driven by streaming fraud, the measure that touches revenue is the second one.
The reliability of any disclosure system depends on who makes the determination. Apple and Spotify start from provider declaration at track level; Deezer decides with its own detection model. Apple has said it has detection technology, but has not said how it uses it in labeling decisions or what happens when a provider declares incorrectly or not at all.
In a 2025 Deezer/Ipsos survey of 9,000 respondents, 80% supported clear labeling of fully AI-generated tracks. In the same survey’s listening test, 97% could not tell fully AI-generated music from human-made music, and 52% said they were uncomfortable with being unable to tell.
From lawsuits to licenses… disclosure becomes a contract-compliance record
Mandatory disclosure arrives alongside a shift in the relationship between AI music companies and record companies. After the RIAA sued Suno and Udio on behalf of the three majors in June 2024, Universal Music Group settled with Udio in October 2025 and Warner Music Group settled with both Udio and Suno in November 2025, each paired with a licensing deal. Forbes reported per-generation royalties of $0.002 to $0.005 (about 2.8–7.0 Korean won) in the UMG–Udio settlement.
AI music generators: litigation and licensing status
As of August 2026. Source: Music Business Worldwide, Forbes, Chartlex
The products changed with the deals. Udio became a closed environment where users can prompt and remix licensed catalog but cannot export, and Suno agreed to replace its existing models with licensed ones in 2026. As licensed generative output enters distribution, the metadata identifying how and under what terms a track was made becomes the basis for royalty accounting and contract verification. The American Federation of Musicians (AFM) suit filed against Warner and Universal in July, alleging settlement money did not reach performers, is what a dispute looks like when that record is unsettled.
Korean platforms block at intake; overseas services admit and disclose
Melon (멜론), Genie Music (지니뮤직) and FLO (플로) have operated so that finished AI tracks produced from prompts alone, without human creative involvement, are likely to be rejected at distribution review. The Korea Music Copyright Association (한국음악저작권협회, KOMCA) does not register works created predominantly by AI. That is the opposite starting point from Apple, Spotify and Deezer, which admit the content and manage it through disclosure.
Korea legislated the disclosure duty first. The Framework Act on the Development of Artificial Intelligence and Establishment of a Foundation for Trust (인공지능 기본법) took effect with its enforcement decree on January 22, 2026, giving legal force to a duty to mark generative AI output. For ordinary output a provider may choose between a human-perceptible mark and a machine-readable one such as a watermark or metadata; violations can draw a corrective order and a fine of up to 30 million won. A grace period of at least a year makes fines unlikely before 2027.
Disclosure frameworks compared: overseas services and Korea
Compiled from company announcements, Ministry of Science and ICT guidance and related commentary
The duty is in force; the track-level field is not built
Korean platforms have no settled track-level disclosure field or on-screen label specification. Apple’s four tags and the RIAA/IFPI two-tier labels are already fields Korean distributors must fill when supplying overseas services. Building a separate domestic specification would mean declaring the same track twice.
For agencies and distributors, an internal standard comes first. Without one, decisions on vocal tuning, synthetic vocals, demo composition, cover art and music-video post-production will vary by staffer, because there is no shared definition of a “material portion.” Apple leaves that judgment to the provider, and the liability for a missing tag stays there.
Virtual artists are judged on identity, not method… the profile decides exposure
Spotify’s AI Persona label turns on whether the artist identity is a photorealistic AI-generated one, not on how the music was made, and a labeled profile is excluded from editorial and algorithmic recommendations by default. For Korean virtual idols and virtual artists, how the boundary between a character-based identity and a photorealistic AI identity is presented in the artist name and profile imagery has a direct effect on exposure overseas.
For OSTs and background music, the production record determines the distribution tag
Broadcasters and production companies commission drama OSTs and variety-show music, contract with performers, and hold archives of past recordings and footage. OST recordings are distributed to domestic and overseas services after broadcast, and by the time they reach Apple Music the track, composition and artwork tags are required. If generative AI use is not recorded during production, it is hard to reconstruct at distribution. Background music and signature themes are where generative AI tools have spread fastest, which widens the range to check.
Performance credits are the other open item. The RIAA/IFPI “AI-Assisted” label assumes a human performed the core creative work with AI contributing expressive elements; how that distinction maps onto agreements under the Korean Broadcasting Performers’ Rights Association (한국방송실연자권리협회, KoBPRA) framework has not been settled. Licensing an archive to a licensed AI model rests on the same terms: the UMG and Warner settlements are built on catalog licenses with per-generation accounting, so without performer consent for reuse and a revenue-sharing basis, disputes of the AFM kind follow.
Sources
· Ethan Millman, “Apple Music to Launch Labels on AI Tracks Later This Year,” The Hollywood Reporter, August 20, 2026
· Kristin Robinson, “Apple Music to Add Labels for AI-Generated Music Later This Year,” Billboard, August 20, 2026
· “Apple Music to Label AI-Made Tracks Later This Year,” Variety, August 20, 2026
· “Apple Music to Label AI-Generated Songs,” MacRumors, August 20, 2026 (Schusser remarks quoted from Billboard)
· “Apple Music launches AI transparency tags,” Music Business Worldwide, March 4, 2026
· “Introducing a New Label for AI-Generated Artist Identities on Spotify,” Spotify Newsroom, August 11, 2026
· Sarah Perez, “Spotify will label ‘AI Persona’ profiles and exclude their music from recommendations,” TechCrunch, August 11, 2026
· “RIAA Pushes for AI Labels on Streaming Services,” The Hollywood Reporter, and “Record industry proposes AI labeling system,” Music Business Worldwide, both July 10, 2026 (first reported by The Wall Street Journal, July 10)
· “RIAA, IFPI & More Come Together to Create System for Tagging AI-Generated Songs,” Billboard, July 10, 2026
· Deezer/Ipsos survey of 9,000 respondents (2025), as cited by iMusician, July 2026
· Virginie Berger, “Launch, Train, Settle,” Forbes, December 18, 2025 (per-generation royalty range)
· “Warner Music Group strikes ‘landmark’ deal with Suno,” Music Business Worldwide, November 25, 2025
· “Music Industry AI Lawsuits Tracker 2026,” Chartlex, updated August 20, 2026 (AFM amended complaint filed July 27, 2026)
· Framework Act on Artificial Intelligence (인공지능 기본법) and enforcement decree, in force January 22, 2026 — Ministry of Science and ICT (과학기술정보통신부) guidance; Bonhwa Law Office commentary, June 2026
· “2026 AI music distribution rules,” RouteNote Korea blog, June 16, 2026 (Melon, Genie, FLO and KOMCA policy)
Quotation practice: double quotation marks indicate remarks as reported; unquoted statements are summaries of reporting. Apple Music’s email language is quoted at second hand from The Hollywood Reporter, Billboard and Variety. Conversion rate used: USD 1 = KRW 1,390.