Roughly four or five out of every ten songs uploaded to Deezer each day are now AI-generated. But barely 3% of what people actually listen to is AI music — and most of even that gets flagged as fraud.
Is AI really "taking over" music?
Deezer's own numbers show 44% of songs uploaded daily are AI-generated. That's roughly 75,000 tracks a day, over 2 million a month. Back in January 2025, it was around 10,000 a day — more than a sevenfold jump in about 18 months.
It sounds like AI is swallowing the music industry. But Deezer's other numbers tell a different story. AI music accounts for only 1-3% of actual streams. Tracks that make up 44% of uploads barely register with real listeners.
Here's the kicker: of that already-tiny 1-3% AI stream share, 85% gets flagged as fraudulent and demonetized. In other words, the share of AI music that's actually being legitimately consumed rounds down to close to zero.
So who's actually behind all this volume?
The case of Michael Smith, a 54-year-old from North Carolina, shows exactly who's generating the spam.
Between 2017 and 2024, Smith generated hundreds of thousands of fake songs with AI and used thousands of bot accounts to stream them on repeat. Instead of hammering one track, he spread plays thin across thousands of songs to dodge fraud detection. Cycling through Apple Music, Spotify, Amazon Music, and YouTube Music, he collected $8,091,843.64 in royalties before pleading guilty to wire fraud conspiracy and forfeiting the entire sum.
What makes this case matter isn't that Smith was an outlier — it's that he was a template. Tools like Suno, Udio, and Google's Lyria made it possible to mass-produce tracks, and that alone removed the barrier to running this scheme at scale.
What Suno is actually changing
Suno sits at the center of this. On August 6, 2026, CEO Mikey Shulman announced three moves in a blog post.
| Before | After the announcement | |
|---|---|---|
| Tracing a song's origin | Nearly impossible on other platforms | Audio watermarking + fingerprinting lets partner platforms identify Suno tracks |
| Mass distribution | No limits | New download limits curb bulk distribution to streaming services |
| Copyrighted lyrics | Reactive, after the fact | Musixmatch partnership screens lyrics at upload |
| Community rules | Vague | Explicit bans on fraud, spam, fake streaming, unauthorized material |
Shulman described the watermarking and fingerprinting as "durable, tamper-resistant, and designed not to affect the listening experience". Suno is also partnering with Audible Magic to screen uploaded audio against copyright databases.
The timing doesn't look like pure goodwill. Universal Music Group and Sony Music are suing over unauthorized training data, and a Munich court sided with German rights group GEMA in July. A November 2025 data breach exposing 55 million users' data has triggered a class action in Massachusetts. Warner Music got ahead of all this by signing a license deal back in 2025. The watermark is both a product feature and evidence Suno can point to in court. Shulman also said it should be up to "artists and platforms" to decide whether to disclose a song was AI-made.
What to check now if your brand uses AI music
This isn't just a musician's problem. If you're a marketer or creator using Suno or Udio for ad music, YouTube content, or podcast intros, this is a good moment to check your setup.
- Check your plan tier
Songs made on Suno's free tier don't carry commercial rights. Upgrading later doesn't retroactively cover what you already made — generate commercial tracks on a paid plan from day one. - Reread the watermark policy
Check how the new watermarking shows up in ads or content you distribute. "No effect on listening experience" doesn't mean no effect on how it's labeled downstream. - Avoid mass-generation patterns
Cranking out dozens of variations and blasting them across channels is exactly the pattern platforms are now built to catch. Fewer, more deliberate tracks is the safer play. - Decide your disclosure stance now
Shulman says it's up to artists and platforms whether to disclose AI use. The same applies to brand content — deciding your policy ahead of time avoids a trust problem later. - Watch for platform policy shifts
Deezer and Spotify are both tightening AI content filters. Legitimate uploads can get caught as false positives, so check your content's status periodically after publishing.
