Deezer: Over 50% of daily uploads are AI-generated

Deezer, the French music streaming platform, revealed that more than 50% of daily uploads in June were generated by artificial intelligence. That amounts to about 90,000 tracks per day — a volume signaling a radical shift in the audio production chain. The company is now racing to adjust its moderation and curation systems.

The technical impact is immediate: Machine Learning algorithms must identify synthetic generation patterns in audio, metadata, and upload behavior. AI detection tools are becoming as essential as anti-fraud systems. Deezer already uses classification models based on spectrograms and convolutional neural networks to flag artificial content, but the false positive rate remains an operational bottleneck.

Business reading

For content platforms, the explosive growth of AI-generated uploads creates a dilemma: scale curation without stifling diversity. From a growth standpoint, moderation automation is the new quality funnel — those who fail to solve this will lose paying user retention. Engagement data shows that playlists with high density of synthetic tracks drive lower listening time and higher churn. The solution lies in using AI to moderate AI, combining classification models with collaborative filtering based on human behavior.

10Dobro view

At 10Dobro Prod, we see this movement as an inflection point for AI systems applied to scale businesses. The ability to operate automation with quality — whether in audio, text, or video — defines competitive advantage in the coming years. For example, in audiovisual, generating hundreds of ad variations on demand requires the same kind of quality control that Deezer faces. What is at stake is not the technology, but the orchestration of processes that ensure performance without losing control over the final outcome.