Andrew Southworth
Videos00:14:11

Spotify Secretly Killed AI Music 5 Months Ago (here's the proof)

Discover how Spotify quietly eliminated algorithmic support for fully AI-generated music months ago, reshaping the platform’s approach to AI versus human artistry.

Quick summary

Spotify has drastically reduced algorithmic exposure for music that is entirely AI-generated, effectively removing it from key discovery channels. Multiple case studies reveal steep declines in algorithm-driven streams for AI music, while human-created tracks maintain or grow their presence. This shift appears intentional, prioritizing human artistry over automated creations within Spotify’s recommendation system. The platform evaluates AI content on a release-by-release basis rather than penalizing entire artist profiles, allowing mixed human and AI projects to coexist with differentiated treatment. This move sparks debate about artistic value, with many supporting the focus on genuine human effort, while others argue for the legitimacy of AI-assisted creativity. Ultimately, Spotify is drawing a clear line between quick AI-generated “slop” and thoughtfully crafted music.

Auto-transcript(English)

Recently, there have been a ton of updates when it comes to AI music in the music industry, but also on Spotify specifically. Spotify came out with this AI persona policy where they're automatically labeling personas or profiles on Spotify that are entirely AI generated. So, the the identity is a fake AI generated persona. They will be labeled as an AI persona and they will be removed from the algorithm and editorial coverage, which I've covered in this video and also an interview with Sam Duoff from Spotify. But what no one has really talked about, and Spotify hasn't talked about it either, is back in around like March, April, uh Spotify 2026, by the way, I'm filming this in early September 2026. So about 5 months ago, Spotify kind of completely killed AI music already. And I'm not talking about like AI personas, but AI generated music. And I have receipts to back it up that I want to show you cuz there's been a lot of discourse on both sides. for example, for human artists saying, "I wish Spotify's policy would go further and that they would not just restrict AI personas, but anyone uploading AI music from algorithm traction." And then the AI people arguing that, hey, I've been seeing my stuff getting restricted anyways, but it but I'm not using an AI persona, but I am generating AI music. So, there's been arguments pushing back and forth from each side and um I'm just here to show you what I've seen and I'll talk a little bit about what what I think is right or wrong at the end of it. But first, let me show you some actual case studies inside of Spotify for artists. [music] >> All right, so here we are inside of Spotify for artists and I'm looking at an artist that has pretty big numbers, right? This is 32 million streams in the last 12 months. And the pink line is algorithm algorithmic traction and the blue line is active traction. So think of the blue as like your ads, your social media marketing, but in this case it's really meta ads driving traffic. And then the pink is what the algorithm is doing to push out the music. So, at its peak, this particular artist was getting 323,000. Let me zoom in for you. 323,000 streams a day from predominantly the radio algorithm. So, when people say that on Spotify they've been hearing AI generated music in all their recommended algorithmic playlists, this is what they're talking about, right? This artist was getting 323,000 streams per day from the algorithm alone. And then comes February 28th and 27th. Um, and fast forward to today, this artist only gets two,500 streams a day. So, we're we're talking from a reduction from 323,000 streams a day down to 1,400. So, it's it's not a 90% reduction, it's a 99 5% reduction in algorithmic traction. Pretty dramatic. And and the first jump happened in the course of like 2 days. It went from 300 down to 55 and then from 55 down to 5 and then it slowly fell from 5 down to 1500. So pretty aggressive purge in this case, right? And if this was just one sample, you'd think you could think a lot of things, right? But in this case, you're seeing the active. So you're seeing their ads didn't really change. They're still pushing this the music pretty hard. Um, and they even released new music here, hoping that it might cause the algorithm to come back. And another one here. But we're going to show more here. So, if I go over to the next example I have here, um, this one, same kind of thing. Getting pretty good algorithmic traction. They're getting 10,000 streams a day. Not as crazy as the other one, right? But still 10,000 streams a day. That's that's no joke from from the algorithm. This this artist got 1.5 million streams in the last year. And keep in mind, I don't think I formally said this, but this AI generated 100% AI generated music. This all the examples I'm showing you until I say otherwise are 100% AI generated music. So this one is 10,000 streams a day and then around April um well first around February it kind of fell near the end, but then by like April 9th it was pretty much dead. So basically in this case from a peak of 10,000 streams a day down to less than a 100 streams a day going forward. Also 100% AI generated artist here um they were having a peak of about 8,000 algorithmic streams a day. Um fast forward to today less than 20 algorithmic streams per day. So it completely died off. And again the blue line like the these artists do still have active listeners right? Um they're they're getting people listening to their music. In the case of this one, 20,000 streams a day from their active audience. So, they do actually have people listen to them on purpose, right? Um, so it's not a case of the promotion stopped. In a lot of cases, the you're seeing this blue line fall, but that's because when the algorithm fell, the the the uh creator, I'm just saying artist just to be generic, but the the creator/ artist of this AI music stopped promoting it as much because the we were like, what is what is going on here? Right? So, um, this is that. And by the way, this these aren't my music projects. These are these are I run an ad agency. We will promote anything if we think it can work. We don't like discriminate versus how the music is made. Um, I leave that up to people who are, you know, clicking on it, listening to it, if they want to listen to it or not. But we predominantly work with human artists and um, you know, I am a human artist, so I appreciate human music. But just want you to know these are not my music projects that I'm showing you here. But this is dramatic, right? These are three case studies. Now, I'm going to draw in another one. All right. I said to pull another window cuz this one's in a different Spotify artist account. But this artist has gotten uh over 10 million streams in the last year. Um and again, pink algorithm blue active. The peak of it was 58,000 algorithmic streams per day. Fast forward to today, less than a thousand. So, a reduction of again um a lot, right? I What is that? 98% reduction in algorithmic streams in this case the same period right around by April algorithm mostly dead in this case the artist got like a second wind around April that then completely died off in July so interesting every other case though it's been around this April 1st um and then another example here now this next example I want to show you it actually makes a big point in this and this kind of proves that a few things about how Spotify works so there's this interesting example I have where it's an it's an artist that has some music that's 100% adenerated and some music that's 100% human generated all released under the same artist project. So I was curious about how this would happen. What's going to happen if an artist is part part AI part human? So this is the AI release. Um and this we're not looking at the whole artist profile. We're looking at just this album here that's AI generated. And around again around April completely dies off. Now, this one didn't have a high peak. It had 1,200 streams per day from the algorithm at its peak and now it's zero, effectively completely zero across the board. We look at the human release and this one is the opposite, right? This one had almost nothing before then and it's been growing. So, now around August, it's getting 17,000 streams a day at its peak algorithmic. So, it's the opposite. And the reason why I think this is a powerful example is it shows that Spotify through through all of these artists I just showed you as far back as March and April have been completely just purging the algorithmic traction from music that is 100% AI generated. I I don't have any music that's like partially AI generated. So all the these examples like this is human generated. Everything else I've shown you has been completely AI generated. It's not like there's real vocals and instrumental AI or vice versa. It's all just like pseudo maybe mastered or something and then thrown online. Maybe lyrics generated. I don't know the details on that, but I know it's pretty much just ad generated from scratch and in like a sunno or something. Um, and it looks like they're kind of they're doing what they said they wanted to do, which is use the algorithm to prioritize human artist. Right? This is a human artist or this case a human album and the other one is a AI album. both in the same artist slot. So, the music is marketed in effectively the same way. It's the same genre of music. It's the same I don't want to say it's the same quality of music because one's human generated, one's AI, but in this case, it's it is um it's just kind of it's instrumental like relaxing piano stuff. So, the quality is uh very similar in this case because of the type of music that it is and the algorithms have grown dramatically since then. So, what does this mean? This means a few things. One, it means that the AI detection isn't necessarily just labeled at the profile level. It's not like if you have one AI release, the whole profile is doomed. It's done release by release or maybe even song by song. In this case, I'm looking at the releases because that's how it was broken up into when the music was distributed, like wholly gener AI generated music, wholly human generated. Um, it seems to not like kill the profile if one is AI generated and the rest is human. Each release is evaluated on its own. The next thing is it does kind of seem like Spotify is like they've crushed the a the algorithmic from AI music and they're giving it to human artists, which is where I think a lot of discussion is going to be in the comments of this video. I have a feeling what you're going to see in the comments below is one human artists being just celebrating this and saying this is the right move. They should be pushing the A music down. And honestly, a lot of artists, a lot of human artists would say that they shouldn't even allow it on the platform. And then on the on the other side, there's going to be the AI artists who are saying that this should be illegal. This is like artificially squashing a different form of music creation arbitrarily. And I think there's some nuance in the middle. I don't know if they're doing this to music that is partially AI generated. I have met people who have done really cool things with AI music. It's not just typing in a prompt and generating it online. That's what all the examples I showed you are. But I have met people where they're getting like demos they recorded from scratch with real instruments, running it through Sunno, coming up with things, splitting the stems, exporting that, putting that in their DAR alongside the rest of their session, and then doing a bunch of post-processing and having this cool thing that like would have never existed 5 years ago. That to me is art. You could argue that yes, it's stolen off the backs of this pseudo thing that's trained without human arts permission. Yes, whatever. Right? That's a thing. That's a separate issue. But that is art. That is someone making something intentionally. It's not the same as slop, which is just cranking stuff out that's entirely adenerated. So separate that out of this discussion because I I think there's a lot more nuance in that area. In my opinion, this is the right move when it comes to the slop. When it comes to just 100% adened music that was one prompt or very few prompt and maybe the songs generated in a minute, 10 minutes, half an hour, whatever. it it doesn't really hold a candle to the amount of time that human artists put in their music, which is always measured in at least hours, but it's pretty much it's very often days, weeks, months that artists are obsessing over their music. Um, and there's lowquality human music that I would argue is also slop that maybe should be in this boat as well, right? I don't know if this should necessarily just be about the fact it's AI generated, but anything that you could genuinely just kind of describe as slop is really the problem. It's it's like slop because it can be updone as scale, whether it's just low effort human music or low effort AI is kind of making it hard for actual artists to make their art and have a sustainable thing. And I think it's a good move for Spotify to prioritize that because that's what the music industry is all about. at least is what it should be in my opinion. And so, um, you know, if you're an AI artist and you're you're very upset about this this algorithmic change, I I get it because you were getting something and now you're not. But I think this compromise is that it allows people to still who are doing who who maybe just want to share the AI music they've generated with the world, throw it on Spotify, still get paid if they're generating real fans, but not have it take anything away from human artists who are putting way more time and effort into their music. Now, you would have actually learned about this like a week earlier if you were part of my newsletter, which is free and you can find to join in the description. But I put a really cool quote on here that I'm just going to read for you because it's an easy way for me to promote my newsletter. It also means if you were only doing this to make money, this just got much harder. There are still plenty of ways to promote and make money with AI music, which I won't be talking about publicly for obvious reasons. And there are plenty of loopholes even on Spotify, but all these changes slowed this stuff down dramatically. Think about it this way. The locks on your front door or car only keep honest people honest. Every lock can be broken unpicked or broken or picked. If someone wants to get in, they're going to get in. But that lock stops 99% of people from trying. Just like how 99% of AI artists who were only in this for the money are probably going to move on to some other get-richqu scheme. The people who were making AI music because they enjoyed it or they found some satisfaction of it or maybe they had some cool art workflow that you could genuinely call art. They don't get really squashed by this. They're going to keep doing their thing. But if you were just in this for the wrong reasons, which is my opinion, just to throw up a bunch of slop and make money, you can't really do that anymore. Where you can, but it just got way harder and you you'd have to learn a lot more to find all these loopholes that I'm talking about. But anyways, if you want to learn how you can promote your music using meta ads, check out this really in-depth tutorial that I have here to see that entire process from start to finish. I'm really interested in hearing what people say in the comments. I am open to having my mind changed if you want to debate it, but make sure you subscribe if you want to see more videos going forward. Anyways, thanks for watching and I'll see you in the next one. Bye.

Grow your music with a proven ad system

Learn the Meta ads and streaming growth playbooks Andrew uses across artists, labels, and agency campaigns.

Explore the course

Keep exploring

Continue in another format

Explore the topicSpotify Growth