Behind this image hides a sniper war for traffic between platforms and bots.
Today is April 22, 2026, exactly Earth Day. Big tech companies are publishing green PR releases, but I’m sitting in front of my computer looking at an outrageous set of data about protecting the “digital environment.”
As a Build-Girl who gets dizzy looking at code all day, I almost spit my tea all over the screen.
75,000 Songs a Day—Who on Earth is Listening to All This?
Let’s drop a bombshell first. French music streaming platform Deezer laid its cards on the table today. Their backend is currently flooded with 75,000 purely AI-generated songs every 24 hours.
75,000 songs.
What does this mean? It means 44% of the daily new content across the platform is spat out by machines. According to public data, this number was merely 10,000 early last year.
Fierce, indeed.
Guess what? Faced with this crazy explosion of productivity, the platform is actually on the verge of collapsing. It’s like opening an independent bookstore, rolling up the shutter in the morning, and finding a truck has dumped tens of thousands of mass-produced, cookie-cutter novels at your door overnight.
There isn’t even room to step.
Tools like Suno and Udio are certainly powerful, driving the marginal cost of songwriting infinitely close to zero. The era of releasing an album just by moving your lips sounds beautiful. But how could platforms possibly accept everything blindly?
A Perpetual Motion Machine: Written by Machines, Listened to by Machines
Here comes the problem.
Are there really that many human listeners consuming these AI-patched miracle tracks every day?
It’s not that simple.
Deezer confessed that although AI songs make up nearly half of the uploads, through strict monitoring and review mechanisms, their actual playback volume across the platform is forcefully suppressed to between 1% and 3%.
Interestingly, up to 85% of this 3% playback volume is identified as fraudulent viewbotting.
So what are these people after?
To be blunt, this isn’t some grassroots artistic renaissance. It’s an extremely secretive underground monetization industry. These gray-market players use AI to mass-produce songs and then use bot accounts to play them on an endless loop, all to extract copyright royalties from the platform. By the way, if someone goes to such great lengths just for a few cents in royalties, I honestly think they’d be better off taking on a couple of freelance gigs.
It’s like someone created an absurd assembly line: machines planting radishes, machines pulling the radishes themselves, and finally taking the bowl to snatch away real human money.
It’s truly outrageous.
Faced with this parasitic operation, Deezer is being honest. They didn’t choose to turn a blind eye. The platform directly deployed a textbook-level physical severing: Step one, algorithmic deranking. Step two, forcibly cutting off monetization channels. They won’t even provide server storage space for high-resolution formats.
To use a colloquial phrase from my hometown in Shandong, this is pulling the firewood from under the cauldron. You’re just after the money, right? I simply won’t give you a single cent.
Honestly, looking at a screen full of red AI upload alerts every day, the review engineers probably want to quit on the spot.
From “Turf Wars” to “Sweeping Trash”
Let’s do a horizontal comparison.
In the past, whether it was Spotify or Apple Music, everyone was competing to see who had the larger copyright library. They would constantly boast about “a 100-million song library” in their ads. It seemed like the more songs you had, the better the platform was.
To be blunt, who dares to brag about massive data volumes nowadays?
When producing a flawless pop earworm only takes ten seconds, “massive volume” itself becomes a burden. If they don’t actively filter it like Deezer does, the recommendation system’s user experience will completely collapse within half a month.
But there’s a catch you should know. Defeating magic with magic comes at a price.
To pick out this industrial waste, Deezer had to develop its own AI detection model. Burning computing power every day to identify another pile of waste generated by computing power—the image is full of dark humor.
I heard they even plan to license this detection tool to their peers. This is probably a side hustle forced upon them.
Sometimes I Wonder: Will This Cat-and-Mouse Game Ever End?
Truth be told, I quite understand the platform’s approach. But there’s one thing I haven’t quite figured out yet.
If one day, a creator uses AI assistance to write a song that genuinely moves me to tears, will it also be deranked and banned?
Exactly how to precisely define the boundary between “assisted creation” and “pure industrial waste” is beyond my realm of knowledge.
A while ago, I went to a Livehouse show. The young guitarist broke a string during an interlude. But he bit the bullet and used the remaining five strings to improvise a killer solo. That sweat-drenched, raw, unpolished feeling is something current algorithms can never compute.
I’ve digressed. Let’s pull back to Deezer’s recent moves.
My superficial understanding is that this is actually an early warning for the entire industry. Future content platforms will absolutely no longer compete on capacity. What they will fight tooth and nail over is whether they can accurately keep soulless waste out of their gates.
When the flood overtops the dam, precise filtering is the final brick.
Casual Thoughts
I’ve been pondering these things for the past few days, and I’ve completely lost the mood to write the code I have on hand.
The evening breeze in Shandong has been quite warm these past two days, though the wind always carries a bit of dry sand.
Closing my notebook.
Time to wash my face and move the succulents indoors.
References:
- Streaming Service Deezer Says 75,000 AI Songs Are Now Uploaded Daily
- Deezer says AI song uploads have nearly overtaken human music | The Verge
—— Lyra Celest @ Turbulence τ.
