Pre-upload quality control for distributors. We catch AI music from unlicensed studios, duplicate uploads, metadata fraud, and recording-fingerprint identity fraud — one API call before your tracks reach Spotify, Apple, or Deezer. Avoid the litigation time bomb that's hitting distributors who don't filter AI content.
The problem
AI-generated tracks uploaded to Deezer every day. 44% of all new uploads.
Growth in AI music uploads in 16 months. Accelerating.
Spotify, Apple, and Deezer are pulling catalog and withholding payments from distributors who deliver AI content.
Your ingest has no pre-upload filter. By the time a DSP catches an AI track, your catalog is flagged, your payouts are held, and your relationship with the platform is damaged.
What's at stake
AI music doesn't trigger blanket DSP penalties on distributors — but it changes the economics. DSPs are demonetizing AI streams. Believe and TuneCore block proactively. Some indie distributors ship AI but withhold Content ID to avoid disputed royalty claims.
DistroShield gives you the information to decide per-track: which generator produced it, whether it's licensed (Udio/ElevenLabs) or unlicensed (Suno still in litigation with UMG/Sony), and whether to distribute, withhold Content ID, or block. The policy stays yours.
The solution
Integrate in 15 minutes. Works alongside any standard distribution pipeline or custom ingest.
POST audio URL + metadata to our API from your ingest pipeline.
AI probability (0–100), classification, confidence, and DDEX-ready tags for Apple's Transparency requirements.
Pass, flag for review, or block — at the threshold you set. Webhooks notify your queue.
POST /v1/analyze
Authorization: Bearer ds_...
{
"audio_url": "https://your-cdn/track.wav",
"metadata": { "title": "...", "artist": "...", "isrc": "..." }
}
→ {
"ai_score": 0.87,
"classification": "ai",
"confidence": 0.92,
"recommendation": "block"
}
What's in the response
Every call runs four independent checks in parallel. Your integration decides what to do with each — pass, review, or block — from a single combined recommendation.
Fine-tuned wav2vec2 analyzes the audio directly — real vocal, synthetic vocal, or AI instrumental. Trained on real distributor catalogs + modern Suno / Udio / MusicGen. Returns generator attribution: Suno (Warner-licensed since Nov 2025; UMG/Sony pending), Udio / ElevenLabs (major-label deals), or unknown (MusicGen/Stable/Riffusion). Distribution and Content ID decisions stay with your policy.
Cross-references ISRC and title+artist against Spotify, Deezer, YouTube in parallel. Catches covers, re-uploads, and tracks already live on DSPs before you distribute a collision.
ISRC format + DDEX completeness + identity fraud (client claims an ISRC that belongs to another artist on Spotify) + artist impersonation + YouTube Content ID fraud (short-track abuse pattern).
Audio-hash match against 70M+ commercial recordings. Catches what string-match modules can't: cross-distributor identity fraud — the same recording registered under multiple artist names on different distributors. The highest-confidence form of fraud.
See the API docs for the full response shape and per-module detail. Or read two real fraud catches from production — anonymized.
Generators detected
Our model is trained and validated on real audio from the AI generators flooding ingest pipelines today. Coverage extends beyond the trained set through the wav2vec2 acoustic backbone — new architectures get caught by family.
Versions through v4. Vocal + instrumental.
Meta's open-source model. Small / Medium / Large.
Same architecture family as Suno.
Stability AI. Caught via backbone.
Spectrogram-based generation.
Loop-based and template generators.
New generators are added to training as they emerge in client pipelines — production review verdicts feed the next model version directly. Per-generator performance is shared in private benchmarks during evaluation. See methodology in docs.
Apple Transparency Tags
Apple Music announced AI Transparency Tags in early 2026 — disclosure metadata for artwork, audio, composition, and music video. Today they're elective. Soon they're mandatory, and the distributor is the party responsible for setting them in DDEX.
Most detection tools stop at a score. DistroShield is built so the detector's verdict and the disclosure metadata can be emitted in a single payload — when Apple finalizes the spec, your pipeline is already wired for it.
One API call returns the AI verdict in a structure designed to ship the DDEX disclosure fields alongside it — no second integration when the policy hardens.
Disclosure field names and value sets land in the response as Apple Music's transparency taxonomy is finalized — mapped to DDEX 4.3, no guesswork on your side.
Spotify and Deezer disclosure schemas plug into the same response shape as their specs publish — one integration, multi-DSP coverage.
Why not a DSP's detector?
Deezer now licenses its internal AI detector to third parties — but that solves a DSP-side problem, catching AI that already entered their catalog. Your penalties are already triggered.
Distributors shouldn't buy infrastructure from the same platforms that penalize them. DistroShield is neutral — no conflict of interest.
Standard pipeline integration, DDEX-native output, distributor-scale pricing — not enterprise DSP licensing.
Apple Music AI Transparency Tags, Spotify disclosure, NO FAKES Act, EU AI Act — handled in one API call.
Blocks YouTube Content ID rejection, Spotify anti-spam flagging, and Deezer demonetization — all pre-upload.
Why us
Our first pilot customer is Mh Musik, LLC — a working distributor processing 400+ releases per month across a 25,000+ track catalog. Every feature is validated against real catalog pressure before it ships to you.
Our detector learns from a blended dataset: verified human tracks from working distributor catalogs, AI samples from Suno, Udio and MusicGen, plus real production uploads labeled by QA teams. Not academic benchmarks.
We know Merlin, FUGA, Revelator, SonoSuite, AudioSalad, CI, DDEX formats, and DSP penalty mechanics from the inside.
Spanish-language support, serving markets other tools ignore.
Pricing
Start with a 7-day free trial — 50 analyses, no card. Pay-as-you-go from there, or pick a plan when volume justifies it. Every tier includes the full detection engine: AI scoring, duplicate check, metadata validation, and recording fingerprint match.
All prices in USD. Pricing may change with 30 days' notice for new billing periods.
Sign up
One-click signup with Google, or fill out the form below. Either way you'll get an API key in seconds. No credit card. Cancel anytime.
Sign up with GoogleYour trial is active. We've also emailed your API key to you for safekeeping.
⚠ This key won't be shown again. We don't store it in plaintext — if you lose it, you'll need to regenerate.
curl -X POST https://api.distroshield.com/v1/analyze \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"audio_url":"https://your-cdn.com/track.mp3"}'