From mid-September, some Spotify artist profiles will carry a badge telling you the artist is not a real person and mobile will get it first.
Profiles carrying the badge will, by default, be left out of Spotify’s editorial and algorithmic recommendations.
On a platform where almost nobody browses a catalogue of hundreds of millions of tracks by hand, that is the whole story.
Two labels, and only one of them is voluntary
From 11 August 2026, artists can go into Spotify for Artists and declare themselves. Do that and the profile gets an AI Persona label.
Don’t do it, and Spotify can apply a Likely AI Persona label after reviewing the account. The company says that review uses a mix of human checks and its own detection tools, and that it will start with artists above certain audience thresholds. Bigger profiles get looked at first.
The badge will show on the artist profile, in the About section, in search results and next to tracks in playlists.
Artists who are flagged can appeal. They cannot pull the label off themselves. Spotify says listeners will eventually be able to report suspicious profiles, but that reporting tool is not there at launch.
The difference between the two labels is worth paying attention to. One means the artist told Spotify. The other means Spotify worked it out.
Using AI tools will not get you badged
This is the bit most likely to be misread.
Spotify is labelling artists who are artificial. It is not labelling music made with AI assistance.
A producer in Katlehong who uses AI to sketch a melody, tidy up a vocal or try a drum pattern is still a real person with a real profile. No badge.
Spotify already has a separate mechanism for that, called AI Credits, which is for disclosing where AI was used in the making of a track.
AI Credits answers whether AI was used. AI Persona answers whether the artist exists.
The short version
- 11 August 2026:Â artists can self identify in Spotify for Artists.
- Mid-September:Â badges start appearing, mobile first.
- AI Persona:Â disclosed by the artist.
- Likely AI Persona:Â applied by Spotify after review.
- The consequence:Â labelled profiles are normally kept out of editorial and algorithmic recommendations.
The badge is cheap. Losing the algorithm is expensive.
Discover Weekly. Release Radar. Editorial playlists. The track that starts playing on its own when the album ends.
That is how most listening actually happens.
Spotify’s product is not access to music. The access is already solved. The product is deciding which sliver of an enormous catalogue gets in front of you.
So being excluded from recommendations is not a cosmetic penalty. The song still exists, you can still search for it, you can still play it. Spotify simply stops handing it to anyone.
Spotify says it wants to prioritise music from authentic artists building careers. Read that as a signal about where streaming is going. Being human is becoming an input to the recommendation system, and anything that affects recommendations affects payouts.
The uploads are enormous. The listening is not.
Deezer is one of the few major streaming services publishing figures on this, so its numbers are the ones everyone quotes. They measure uploads, which is not the same as demand.
In July, Deezer said fully AI-generated music had passed half of everything uploaded to the service each day, peaking at around 90,000 tracks a day in June. At the start of 2025 that figure was closer to 10,000.
Then the other half of the number. AI-generated music accounts for somewhere between one and three percent of actual streams on Deezer, and Deezer says roughly 85% of those streams are fraudulent and get demonetised.
So an enormous amount of music is being made that almost nobody is choosing to listen to.
That points at what a lot of this is for. Upload in bulk, point automated listening at it, extract royalties.
Spotify has been fighting the broader version of the same problem. In September 2025 it said it had removed more than 75 million spammy tracks over the previous 12 months. That number is not 75 million AI songs. It includes duplicates, mass uploads, search manipulation and general junk.
Spotify has not published how much AI-generated music is being uploaded to its platform. Neither has Apple Music or Amazon Music. Until they do, Deezer’s numbers are a proxy, not a measurement of the industry.
South Africa has already run this argument once
Rea Gopane’s Suka!, released in December 2025, was widely described as South Africa’s first AI-assisted amapiano hit. Gopane has acknowledged using AI without saying which tools.
It charted on Apple Music South Africa, which is the part that made it a music industry conversation rather than a tech curiosity.
Cassper Nyovest and Vusi Nova have both spoken about finding out that music they liked was not coming from the human artist they assumed was behind it.
Underneath that sits a harder problem, which is training data. A local data audit reported in July found South African artists appearing repeatedly in documented AI training datasets. Kabza De Small reportedly appeared 360 times, DJ Maphorisa 270, Spirit of Praise 212, Nasty C 193 and Thomas Chauke 183.
Those counts describe how often names show up in documented datasets. They do not, on their own, prove what any particular model was trained on.
The question is not whether a machine can produce an amapiano track. It is what it had to be fed first.
A CISAC study by PMP Strategy estimates that 24% of music creators’ revenues globally could be at risk by 2028 as generative content substitutes for human work. Amapiano is one of South Africa’s largest cultural exports, so that estimate lands here harder than most places.
The royalty admin is where this actually hurts
South African royalties already move through SAMRO, CAPASSO and SAMPRA, and the whole system runs on metadata. Who wrote it, who performed it, who owns what percentage, where it was played.
Get that wrong and the money does not reach the right person.
Now add hundreds of thousands of synthetic tracks with copied names, near identical titles, missing credits and automated plays.
A labelling problem turns into an accounting problem, and an accounting problem turns into an artist problem.
The badge does not fix any of that. It just makes one part of it visible.
A real person releasing synthetic music walks straight through
Suka! is the clearest example of the gap.
Rea Gopane exists. There is a real person behind the profile. So even if the music itself is largely generated, the profile does not obviously qualify as an AI Persona, because Spotify is judging identity rather than output.
The second gap is the threshold. Starting reviews with high audience profiles is sensible if you want to protect the largest number of listeners. It also means smaller AI profiles stay untouched for a while.
If the fraud is spread across thousands of small accounts rather than a few large ones, that is the harder half of the problem and it is not being addressed yet.
Europe wrote a rule. South Africa has not.
Spotify’s announcement came shortly after transparency requirements under Article 50 of the EU AI Act became applicable on 2 August. Those rules cover identifying artificially generated or manipulated content.
The timing may be coincidental. Spotify has not linked the two.
South African copyright law was written long before software could ingest a catalogue and produce something adjacent to it, and the Copyright Amendment Bill is still unresolved.
For now, the rule that affects what a South African musician’s profile looks like is being set in Stockholm, not Pretoria.
What you are actually paying R69.99 for
Spotify’s local pricing has moved around. Standard went up to R94.99 in November 2025, then Spotify dropped Premium Lite in South Africa and brought Standard back down to R69.99 a month.
Student is R37.99. Platinum is R179.99. Basic Platinum is R119.99, which gives you most of Platinum without the audiobook hours.
Spotify bills you directly, so there is no street price to shop around for and no retailer undercutting anyone.
What you are paying for is not the catalogue. Apple Music has the catalogue. YouTube Music has the catalogue. The overlap is close to total.
You are paying for the sorting. The playlists, the recommendations, the track you had never heard that turns out to be good.
Which is why this announcement is larger than it looks. Once a catalogue is effectively infinite, adding to it stops being worth anything, and deciding what is worth playing becomes the product.
What to do about it
If you just listen: from mid-September, check the About section on an artist profile before you assume a new favourite is a person. The badge appears there, in search and in playlists.
If you release music: if your public artist identity is AI generated, self identifying from 11 August gives you the AI Persona label rather than waiting to be flagged with Likely AI Persona. Either way, expect to lose recommendation placement, and plan your release around direct audience rather than Discover Weekly.
If you use AI as a tool and release under your own name, nothing here applies to you. Use AI Credits to disclose it and carry on.
