A six-layer pipeline that takes raw distributor CSVs and certifies every metadata field against Spotify and Apple Music compliance guidelines — across 150+ stores, in any language.
CATCHES: Non-English metadata invisible to downstream layers
OUTPUT: Semantic concept representation
CATCHES: Mix / Remix / Live / Instrumental in Title instead of Version field
OUTPUT: Formatting issues flagged
CATCHES: Bare keywords, stacked descriptors, structural SEO, genre stacks, "for [use]" patterns, frequency terms
OUTPUT: FAIL / REVIEW / PASS
CATCHES: All-functional phrases, genre+venue, persona patterns, weak wrapper rescues
OUTPUT: SEO / BORDERLINE / CLEAN
CATCHES: Instructional patterns ("to make"), topic+format, reconstruction overrides
OUTPUT: Gate escapes caught
CATCHES: Missing ISRCs, duplicate ISRCs, blank Action fields, any layer still failing
OUTPUT: SHIP / NO-SHIP
Gate Clean ≠ Distro Ready — All 6 layers must pass before metadata ships
Real before/after transformations. Each rewrite preserves the thematic spirit of the original while eliminating search-intent descriptors.
Track Title
Music For Kittens To Sleep
Track Title
The Purring Hour
Instructional pattern detected ("for...to sleep"). Rewrite preserves animal theme without SEO.
Preserves pet/animal connection
The classifier asks one question: does this phrase name an artistic entity, or does it describe searchable audio content?
The Core Rule
Metadata must name a real identity
or work — not describe search intent
The decision is made at the phrase level, not the word level. "Jazz" is a genre word — but "Gypsy Jazz Caravan" is a band name.
Relaxing Piano Music
Use-case + instrument + format. Fully reconstructible — a listener can guess the audio from the title alone.
Detected at L2Gypsy Jazz Caravan
"Caravan" is a creative anchor — evokes imagery, Django resonance. Cannot be predicted from audio content.
Detected at L3Deep House Lounge
Genre + venue. Describes a listening environment, not an artistic identity.
Detected at L2Viking Music
"Viking" is a thematic/cultural identifier, not a functional descriptor. Reads as a project identity.
Detected at L3Jazz Ensemble
Weak wrapper ("Ensemble") cannot rescue alone. "Jazz" is the only testable content — it's functional.
Detected at L3Winterbell Ensemble
"Winterbell" is a coined word — a strong identity anchor. Wrapper + strong anchor = plausible entity.
Detected at L3Baby Mozart Music
Topic + format. "Baby" is audience, "Mozart" is topic, "Music" is format — not an identity.
Detected at L4Baby Lullaby Academy
Persona pattern. "Academy" is an institution anchor — reads as a named project, not a description.
Detected at L3Music To Make Your Kitten Calm Down
Instructional pattern ("to make"). Describes the function of the recording, not its identity.
Detected at L4The Queen of Soul
Persona pattern: "[Title] of [Domain]". This is identity, not description, regardless of functional words.
Detected at L3Strong Anchors
Person names, coined words, institutions, verified brands. Can rescue a phrase alone.
Weak Wrappers
Ensemble, Orchestra, Band, Choir, Collective. Cannot rescue alone — need a creative anchor.
Thematic
Viking, Celtic, Ocean, Forest, Sacred. Context-dependent — not functional, not strong identity.
Compliance is determined by semantic meaning — not by language, spelling, or script. Equivalent metadata in any language receives an equivalent verdict.
Same meaning, 5 languages
Relaxing Music
EnglishMusique Relaxante
FrenchMúsica Relajante
Spanish癒しの音楽
JapaneseEntspannungsmusik
GermanNormalized Concept
USE_CASE + FORMAT
All 5 expressions receive the same compliance verdict. The engine reasons over semantic meaning, not vocabulary.
Language-Independent Concept Classes
USE_CASE
relaxing, study, sleep, focus, workout
FORMAT
music, beats, melody, instrumental, mix
WELLNESS
healing, meditation, calming, soothing
GENRE
jazz, rock, techno, lofi, ambient
AUDIENCE
children, baby, kids, bambini, enfants
VENUE
café, restaurant, lounge, bar, club
The deterministic gate (L2) can't catch everything. When SEO values survive the gate, they're treated as structural problems — not keyword problems. Each escape pattern becomes a new rule, a regression case, and a permanent recall test entry.
12
Escape patterns
77/77
Recall tests passing
132
Regression cases
Every content word is functional (genre, format, use-case, venue) with no creative anchor
Study – Deep Focus
Light Cafe
Working & Chill
Streaming Melodies
A genre word + a venue word describes a listening environment, not an identity
Jazz Café Bar
Deep House Lounge
Cappuccino Lofi Lounge
A musical modifier (melodic, deep, smooth) + a genre word = subgenre description
Melodic Techno
Smooth Jazz
Deep House 2024
Electronic Deep
A use-case word (study, gaming, sleep) + a format or venue word = functional description
Study Sessions
Gaming Melodies
Relaxing Piano Music
Soothing Background Music
"to make / to help / to calm" constructions describe the recording's function, not its identity
Music To Make Your Kitten Calm Down & Go To Sleep
A topic word (Mozart, Halloween, Baby) + a format word — the topic is subject matter, not an identity anchor
Baby Mozart Music
Halloween Background Music
Best Country Music For Office
Creative-sounding modifiers don't override a fully reconstructible description
Deep Blue Ocean Meditation for Insomnia Relief
Version words (Mix, Remix, Live, Instrumental) inside the Title field instead of the Version field
Animal Crossing Music Mix
Animal Crossing Radio Mix
Stop losing revenue to metadata rejections. Start validating your catalog against every major music store today.