If you’ve spent enough time scrolling on TikTok, you’ve probably become accustomed to many songs without ever putting them in a playlist — through slideshows, art trends or edits of your favorite show. Listening to music has become something that begins on other platforms, before eventually becoming familiar enough to earn a spot in our libraries.
Finding the artist usually follows. For listeners discovering The Velvet Sundown, that search result led to a band whose members did not exist.
The Velvet Sundown appeared out of thin air in June 2025, releasing “Floating on Echoes” and “Dust and Silence” within mere weeks of each other. Its folk-and-country-inspired sound harkens back to 1960s and 1970s folk and psychedelic rock, bearing a resemblance to classics like Crosby, Stills, Nash & Young. By early July, it managed to amass a following of 1 million monthly listeners on Spotify. With two albums and 26 total songs released, the band appeared to be doing remarkably well for itself in such a short time span.
Listeners soon found that the musicians shown in the promotional photos seemed to be doctored with the use of artificial intelligence. After a slew of criticism and scrutiny from listeners over its AI-generated photos and lack of live performances, the band changed its Spotify description, identifying itself as “a synthetic music project guided by human creative direction,” acknowledging that AI was involved in its instrumentals, vocals and visuals.
Full-scale AI songs and bands are hardly new to the online music scene.
As far back as April 2023, “Heart on My Sleeve” circulated online, featuring vocals generated to resemble Drake and The Weeknd. The track was uploaded by a creator known as Ghostwriter and gained millions of views before it was removed after objections from Universal Music Group. Neither Drake nor The Weeknd were aware of the initial release.
These instances of AI-generated music are alarming but are merely a few examples. Deezer, a global music streaming service, reported 90,000 fully AI-generated tracks uploaded daily in June 2026 — more than half of all new uploads.
In the past, tools like these might have produced simple songs; now anyone can make something that resembles a polished release.
Hamza Zakir, The Velvet Sundown
The sudden influx of AI music is propelled by barriers to entry for music production being lower than ever. Tools like Suno and Udio can produce vocals, lyrics and instrumental tracks just from a single prompt. In the past, tools like these might have produced simple songs; now anyone can make something that resembles a polished release.
While the percentage of AI music released is not proportional to the percent streamed — fully AI-generated music only accounts for 1% to 3% of streams on Deezer — a single percent is still alarming. But how do these tracks and bands gain traction in the first place?
The answer lies both in the way that listeners are increasingly introduced to music online and the fact that many AI-generated tracks are not immediately obvious in their production.
Spotify recommendations and social media introduce people to new music without so much as a look at the band beyond a single song. AI music slips through the cracks of algorithm-generated recommended songs on streaming sites. On social media, people can hear songs dozens of times without even knowing their names or the artists behind them.
A Deezer-commissioned Ipsos survey published in November 2025 asked 9,000 people across eight countries to classify three tracks: two AI-generated songs and one human-made recording. A whopping 97% failed to identify all three correctly. Still, 80% of participants supported clearly labeling fully AI-generated tracks, and 45% wanted an option to filter it out altogether.
These statistics show that even if our ears can no longer tell the difference, people still demand transparency. The desire to label and filter this kind of music represents a desire for human-made art in a digital landscape shaped by algorithms and AI.
In today’s world, it is no longer strange to discover a song before discovering its artist. Now, a listener may put a song on a playlist without knowing whether the singer, band, photos or person attached to it exists at all.