The AI music debate stopped being philosophical this month. It became a metadata problem, and metadata is what distributors deliver.
On July 15, TIDAL flips a switch. Tracks it judges to be wholly or substantially AI-generated get a visible label and lose royalty eligibility, according to Variety. Human-led work that uses AI as a tool still earns. The line between those two outcomes now runs straight through your delivery file.
What actually changed
DSP means digital service provider, the streaming platforms like Spotify, Apple Music and TIDAL. Three of them moved on AI inside twelve months, and they did not move the same way.
- TIDAL labels fully AI tracks and blocks royalties on them from July 15, plus removes content built to manipulate recommendations or payouts.
- Deezer tags AI tracks but pays legitimate streams normally. It now receives roughly 75,000 AI uploads a day, about 44% of all new music, per the Deezer Newsroom.
- Spotify backs a disclosure standard rather than a ban, letting creators specify exactly which parts were AI-made, per its September 2025 policy.
Three platforms, three answers, one shared dependency. All of them read the metadata your distributor sends.
DDEX is the referee now
DDEX means Digital Data Exchange, the industry standard that formats the credit and rights data every distributor delivers to every DSP. Spotify’s approach leans on a new DDEX field for AI disclosure, and because DDEX is near-universal, the same tag can travel to Apple Music, Amazon and YouTube Music later.
That is the quiet part. You do not label your own AI track at TIDAL. Your distributor’s delivery does. If that field is blank, wrong, or absent, a platform’s own detector decides for you, and detectors are blunt.
Deezer reports that up to 85% of streams on fully AI tracks were fraudulent in 2025. Platforms tuning filters against numbers like that will err toward suspicion, not nuance.
Why the fraud case raises the stakes
The enforcement mood is not abstract. On July 29, Michael Smith is sentenced after pleading guilty to using AI-generated songs and roughly 10,000 bot accounts to steal more than 8 million dollars in royalties, the first US criminal case of its kind, per the Department of Justice.
That case is why platforms now treat undisclosed AI plus bulk uploads as a fraud signal, not a genre. Legitimate AI-assisted artists get caught in the same net when their metadata does not clear them.
What indie artists and labels should do before July 15
The actionable version is short.
- Disclose precisely. If AI wrote the topline but you sang and played it, say exactly that. Binary “AI or not” declarations lose you money on Spotify and safety on TIDAL.
- Check your distributor supports the DDEX AI field. If it cannot pass structured AI-service disclosure, your human-led track can be mislabeled at delivery.
- Keep session proof. Stems, project files and co-writer splits are your appeal evidence if a detector flags you wrongly.
- Watch your upload velocity. Bulk drops read as fraud signals now, regardless of intent.
This is where a distributor’s plumbing stops being invisible. A DDEX-native pipeline that captures the AI service used at quality control, rather than bolting on a checkbox, is the difference between a track that earns and one that gets muted at the source. It is the same reason regional catalog owners lean on platforms built to pass clean, structured metadata to every DSP rather than the three that dominate the headlines.
TIDAL is the first to make disclosure a payment condition. It will not be the last. The artists who win the next twelve months are the ones whose metadata already tells the truth, cleanly, before a filter has to guess.
