Deezer: 85% of AI-track streams fraudulent. Use Sume Music for beds

Deezer's newsroom reports that up to 85% of streams on fully AI-generated tracks were identified as fraudulent in 2025 and were demonetized. If your plan was to earn from streams of AI tracks, that is a warning. If your plan is background music inside a video you own, the stream-fraud problem does not apply, and Sume Music at $0.125 per accepted generation is a direct fit.

The numbers

Deezer's own page is the only source for these figures.

Use the music where it earns without streams

Generate an instrumental bed with sume/music-auto, then place it under a video with Timeline 1.0 soundtrack or pre-mix with timeline audio. Duck the bed under voice with soundtrack.duck_db.

Prompts accept up to 5,000 characters and a length cue such as "a 60-second track"; there is no duration field.

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": "bed-001"},
    json={"model": "sume/music-auto",
          "prompt": "A 60-second warm acoustic bed, no vocals, steady tempo."},
)
print(r.status_code)

A cautious note

Do not read this as legal or monetization advice for any platform. Read each platform's own policy before releasing AI music; Deezer's rules apply only to Deezer.

Worked example

Consider two ways to use the same track.

• Streaming release: subject to platform detection, tagging and fraud checks as Deezer describes.
• Video bed: the track plays under your own footage and is judged by the video platform's rules, not a streaming service's.
• Cost per bed: $0.125 per accepted generation; three variants cost $0.375.

Checklist before you commit

Because the fraud figure is a platform statistic, do not infer anything about any single track from it.

• Read the video platform's music policy.
• Keep generated beds instrumental if you do not want lyrics to clear.
• Archive the artifact and job id.