Deezer tags AI music: what to keep from a Sume Music job
Deezer tags fully AI-generated tracks and removes them from recommendations and editorial playlists. What a Sume Music job returns and what record to keep.

Deezer's newsroom says it explicitly tags fully AI-generated music for listeners and removes detected AI tracks from algorithmic recommendations and editorial playlists. If you publish Sume Music output as a track, expect that tag to apply if the detector flags it, and plan for distribution without recommendation reach. Sume does not control Deezer's detection; what it gives you is a clear job record of how the audio was made.
What Deezer says
The figures below are from Deezer's July 2026 newsroom post.
| Topic | What the page says |
|---|---|
| Tagging | AI music explicitly tagged for listeners as of June 2026 |
| Recommendations | Detected AI tracks removed from algorithmic recommendations and editorial playlists |
| Volume | About 90,000 fully AI-generated tracks per day, over 50% of uploads at the June 2026 peak |
| Fraud | Up to 85% of streams on fully AI tracks identified as fraudulent in 2025 and demonetized |
What a Sume Music job gives you
POST /v1/music-router/generate returns an audio artifact (audio/mpeg hosted on media.sume.com) and, when present, result.lyrics. job.request.routed_model names the engine that produced it. Price is $0.125 per accepted generation.
Keep the job id, the prompt and routed_model with the file. If a distributor or platform asks how a track was produced, you can answer from your own records.
import os, requests
r = requests.post(
"https://api.sume.com/v1/music-router/generate",
headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
"Idempotency-Key": "track-ledger-001"},
json={"prompt": "A 90-second ambient piano bed. Instrumental."},
)
job = r.json()
print(job.get("id"), job.get("request", {}).get("routed_model"))Where AI tracks still fit
Background beds for video, podcasts and ads do not depend on streaming-service recommendations. For those uses, the Deezer policy is mostly irrelevant. If your goal is streaming-platform discovery, assume a fully AI track will be labelled.
Worked example
A simple record-keeping routine costs almost nothing.
- Store the job id, prompt and
routed_modelin your asset sheet. - Save the audio artifact itself; do not rely on a hosted URL alone.
- Note the date, since platform policies change.
Checklist before you commit
If you plan a release, read your distributor's AI rules as well as Deezer's. Deezer's page describes its own platform only.
- Decide whether the track is a video bed or a release.
- Tag AI-made tracks in your own catalog.
- Do not claim a human performed what a model generated.
Sources
Related posts
More in Use cases
- Demand Gen dynamic product videos from Merchant Center: a clip per SKU
Google says videos uploaded to Merchant Center can run across Demand Gen based on live interest. Make one short product clip per SKU on Sume and upload each.
- Demand Gen one-click image ads: match the landing page
Google's September 2026 Demand Gen Drop adds one-click landing pages from image ads on YouTube Shorts and Gmail. Make a matching page hero with Sume.
- Diwali greeting video from photos: music bed and Timeline render
Make a 20-second Diwali greeting video from four photos: generate an instrumental bed with Music Router, add fades, and render vertically for status posts.
- Does the EU AI label replace an ad disclosure? Article 50(6) says no
Article 50(6) leaves other transparency law untouched. An AI label and an ad disclosure are two lines, and Sume caption cues can burn in both.
Written by Sume