Deezer Is Drowning in 75,000 Robot Songs a Day — And 85% Are Faking Their Own Plays
Deezer Is Drowning in 75,000 Robot Songs a Day — And 85% Are Faking Their Own PlaysNearly half of all new music hitting one big streaming app was made by a machine. And most of those “listens”? Bots pretending to be you.
75,000 AI songs uploaded PER DAY. That’s 44% of everything new. 2 million+ a month. And 85% of their plays got flagged as straight-up fraud.
OKAY SO the music app Deezer just opened the books on something wild, and honestly it says way more about where the internet is heading than about music. Full breakdown on TechCrunch here.
WAIT. Let me back up, because when I read this I actually said “no chance” out loud (my roommate was concerned).
Deezer is basically a Spotify competitor — a big app where you stream music. They just admitted that 44% of every new song getting uploaded to them is made by AI. Not sung by a person. Not played on a real guitar. Just typed into a robot (“make me a sad country song about a truck”) and boom, it spits out a finished track. Around 75,000 of these a day. Over 2 million a month.
And here’s the part that made me lose it: 85% of the plays those AI songs get are fake. Bots. Fake accounts pressing play over and over to farm money. Deezer caught it and cut off the cash. Wild.
🧩 Dumb Mode Dictionary (read this first, takes 20 seconds)
| Word people throw around | What it actually means |
|---|---|
| AI-generated music | A robot writes and “performs” the whole song. You type a sentence, it makes a track. No humans, no instruments. |
| Streaming platform | An app like Deezer/Spotify where you play music. Artists get paid a tiny bit every time someone listens. |
| Stream farming / fraud | Using fake accounts or bots to press “play” thousands of times so you collect money for listens that never really happened. |
| Demonetized | The platform caught the fakeness and said “nope, you get $0 for those plays.” |
| Royalty pool | The big pot of money split between all artists. More songs = your slice gets thinner. |
| AI tagging/detection | A tool that sniffs out “hey, a robot made this” and slaps a label on it. |
📰 How we got here (the numbers are climbing scary fast)
This didn’t happen overnight — it’s a rocket ship. Straight from Deezer’s own newsroom post:
- Jan 2025: ~10,000 AI tracks a day
- Sept 2025: ~30,000 a day
- Nov 2025: ~50,000 a day
- Jan 2026: ~60,000 a day
- Now: ~75,000 a day — 44% of ALL new uploads
Deezer was the first app to start slapping “made by AI” labels on tracks back in June 2025. In one year they tagged 13.4 million robot songs. Thirteen. Million. Music Business Worldwide has the deep dive.
💰 The receipts (why anyone bothers making 75k robot songs)
Here’s the money logic, dead simple:
- A streaming app pays roughly a fraction of a cent per play.
- One song making pennies = pointless.
- But 100,000 robot songs, each getting a few bot-plays a day? Now you’re skimming real money off the top.
- It costs basically $0 and 10 seconds to make each AI song. So the math is: flood the zone, farm fake plays, cash out.
That’s the scam. And it’s why 85% of AI streams got flagged as fraud — the songs weren’t made for people, they were made for bots to play. Deezer even built a tool that spots AI music on OTHER apps now (Spotify, Apple Music), so this is industry-wide.
🗣️ What the timeline's saying
- Real musicians: furious. Every robot song thins out the money pool they share.
- The “so what” crowd: “if only 1-3% of actual listening is AI, who cares?” (Fair — most humans aren’t choosing robot songs on purpose.)
- The hustlers: already reverse-engineering which genres the bots hide in (lo-fi, sleep sounds, white noise, ambient — stuff nobody listens closely to).
- The realists: this is the same fake-engagement game that broke YouTube and Instagram, just wearing headphones now.
The Hacker News thread got heated — worth a scroll if you like watching engineers argue.
🧠 The bigger thing this quietly proves
Forget music for a sec. The real lesson:
When making a thing costs zero, the internet fills up with that thing until “real” becomes the rare, valuable part.
Robot songs are just the first flood because music is easy to fake and pays per-play. Next it’s AI podcasts, AI product reviews, AI “news” sites. The winners of the next few years aren’t the people making the flood — it’s the people who can prove they’re human, or the people who build the filters that catch the fakes. Keep that in your back pocket.
Cool. Half the new songs are robots and the bots are clapping for themselves… Now What the Hell Do We Do? (ง •̀_•́)ง
Look — I’m not telling you to go farm fake plays (that’s the thing getting demonetized and banned, dummy). The money is in the reaction to the flood, not the flood itself. Here’s where a broke 22-year-old with a laptop actually wins:
🕳️ The Human Stamp Hustle
Every app is now scrambling to prove which artists are actually people (Spotify literally just rolled out “verified human” badges). Small musicians are confused and scared. Be the person who sets up their “proof I’m real” kit — behind-the-scenes studio clips, verified profiles, signed uploads — as a done-for-you package.
Example: A 24-year-old in Lagos, Nigeria noticed local Afrobeat artists getting their real songs wrongly flagged as AI. He started a $40 service on Fiverr setting up their verification + filming a 15-second “here’s me recording it” proof clip. Word spread in one WhatsApp group. ~$900/month, part time.
Timeline: First paying client in ~1 week. Stays good until the apps make verification one-click easy (maybe 6-9 months) — ride it now.
📡 The Fraud-Radar Snitch (grey-hat, totally legal)
Deezer says 85% of AI plays are fraud. That means labels, distributors, and indie artists are TERRIFIED of accidentally paying for fake streams. Build a simple report: pull a song’s public play stats, flag the tells of bot-farming (weird 3am spikes, all-plays-no-saves, one country only), sell it as a “is your money going to bots?” audit.
Example: A 21-year-old data nerd in Manila made a Google Sheet template that eyeballs public streaming analytics for fraud patterns. Sells the template + a 30-min walkthrough call to small labels for $75 a pop on Gumroad. No coding degree, just pattern-spotting.
🎣 Bait the Overwhelmed (the niche cheat-sheet play)
When a new mess appears, the FIRST clear “how do I deal with this” guide becomes the thing everyone links to. Nobody’s written the plain-English “How indie artists survive the AI flood” playbook yet. Write it once. Let Google feed you readers forever.
Example: A bedroom producer in Poland wrote a free 12-page guide “Don’t Get Demonetized: The 2026 Indie Upload Rules” and put it on a one-page site. Free guide pulls people in, a $9 “checklist + templates” pack cashes them out. This is the classic content-to-product funnel — Notion or Carrd is all you need to build it.
🪟 The Sound-Effect Loophole (before everyone catches on)
Here’s a legit angle hiding in plain sight: apps are punishing AI music, but real creators desperately need cheap, cleared background sounds — lo-fi loops, rain, ambient beds for videos and podcasts. AI is genuinely GREAT at that boring-but-needed stuff, and it’s not the fraud they’re hunting. Make packs of AI-generated background audio, sell them to video editors who don’t want copyright headaches.
Example: A student in Vietnam used Suno to generate 50 royalty-free “coffee shop ambient” loops, bundled them, and sells the pack to YouTubers on Gumroad for $12. He’s transparent it’s AI — that’s the selling point (no copyright claims). Repeat buyers every month.
🔍 The Real-Ear Curator (be the taste a robot can't fake)
Machines can make 75,000 songs a day but they can’t have taste. Humans are already exhausted sifting robot slop to find real gems. Become the trusted human filter: a tiny weekly playlist or newsletter of “100% human-made, hand-picked” tracks in one specific vibe. People pay for a curator they trust when everything else is noise.
Example: A 23-year-old in Brazil runs a “Verified Humans Only” indie playlist on a Spotify for Artists profile + a free Substack. Artists pay a small fee to be considered (not guaranteed) for the list; listeners get a trusted source. ~$600/month from submission fees alone, growing.
🛠️ Follow-Up Actions
| If you want to… | Do this today |
|---|---|
| Understand the fraud scale | Read the TechCrunch report |
| See who’s fighting AI music | Check Spotify’s human badges |
| Make legit AI background audio | Play with Suno |
| Launch a guide/product for $0 | Spin up a page on Carrd or Gumroad |
| Sell a service fast | List it on Fiverr |
Quick Hits
| You Want | You Do |
|---|---|
 Grasp the flood |
75k AI songs/day, 44% of new uploads — read it |
 Spot the scam |
85% of AI plays = bot fraud, demonetized |
 Sell “real human” proof |
Package verification for scared indie artists |
 Use AI the legal way |
Make cleared background audio on Suno |
 Be the filter |
Curate a human-only playlist people trust |
When making stuff costs nothing, “made by a real person” becomes the luxury item. Sell that.
!