Listeners struggle to spot AI-made songs as uploads surge on streaming

In short: AI-made music is getting so hard to recognize that most listeners cannot tell it apart from human-made songs, and streaming services say the volume is surging.

What's going on

A growing number of people say they dislike not knowing whether a song was made by a person or by AI. In a 2025 survey tied to Deezer, 97% of listeners failed to correctly identify whether tracks were AI-generated or human-made. More than half said that made them uncomfortable.

The reason this feeling is spreading is simple: AI music is showing up everywhere on streaming apps. Deezer said it detected more than 13.4 million AI tracks in 2025. It also reported that in January 2026 it was seeing about 60,000 AI uploads per day, which it said was 39% of new music delivered to the service.

Some AI music has blended in well enough to build a real audience before anyone noticed. One example mentioned in reporting is an AI “band” called The Velvet Sundown, which reportedly drew over 1 million monthly plays on Spotify before listeners learned it was AI.

Platforms are reacting in different ways. Deezer says it can tag AI music from tools such as Suno and Udio, remove it from recommendations (like taking it out of the “you might also like” shelf), and cut off payments tied to fake listening. Spotify has said it removes spammy uploads and bans unauthorized voice cloning (copying a singer’s voice without permission). YouTube has added labels when it detects AI content.

What to watch

A big open question is whether streaming services can keep AI music transparent and limit fraud. Deezer has said as much as 85% of streams on AI tracks can be fraudulent, compared with 8% overall, which could affect how money is shared with working musicians.

Source: The Verge AI