The Recording Academy just published a panel discussion from a “Grammys on the Hill” event where CEO Harvey Mason Jr. repeated his favorite talking points regarding generative AI music. He said:
“AI is ubiquitous in production of pop and R&B and some genres of hip hop music. It’s everywhere. Period. Every producer, every songwriter, every studio that I’ve been in or people that I’ve worked with, every single one of them is using AI.”
Besides this being patently false (funny, the studio I was in last week didn’t use it; or the myriad of fellow songwriters, producers and artists I’ve spoken with recently – many of whom are Grammy winners of every genre – also, don’t touch it), it makes you wonder, why is Harvey continuing to push this narrative?
He goes on to compare generative AI music to other technological advancements throughout the history of music creation like synths and sampling.
He recalls how the naysayers back in the day would say hip hop and R&B were “not music” because they used samples or synths. So, apparently his takeaway from those lessons is to embrace all new tech no matter what. The irony is by not wanting to seem like an out-of-touch luddite, he is ignoring the blaring alarm bells being rung by AI safety experts, who argue that this technology is, in fact, very different from any technology that has come before. AND, more importantly, he’s ignoring objections from the artist community that he claims to represent who don’t. want. this.
View this post on Instagram
There’s a massive difference between using a synth or drum machine in a production, and typing in a prompt and generating a radio-ready track in seconds. An accountant with a Suno subscription is not the same as Prince using synths and drum machines or Dr. Dre sampling a Parliament track.
There’s innovation with music tech and then there’s abomination.
“And yet for us, the idea of optimizing what we do is a complete miss of the entire point of what compels us in the first place. So to everyone who is gassed up about the new ways you can fake making art, by all means drive right off that cliff. We’re genuinely happy to see you go … godless whores.”
– Jack Antonoff
If everyone is using generative AI music creation tools and it is so ubiquitous, then why is no songwriter, artist or producer (who isn’t being paid by Suno) proudly talking about it? If it actually is used in every pop and R&B room, let’s hear it from them. Jack Antonoff has taken a stand and clearly doesn’t use it. No one’s going to question his resume. But sure, let’s hear from Finneas and Billie Eillish. What say you, SZA? How’s the prompting going Chappell, Bruno, Sabrina, Olivia, Bad Bunny, Doechii, Kendrick, Taylor, Queen Bey? And their less famous co-writers and producers. Well, if it’s true, they’re not saying anything because they’re ashamed and embarrassed. And if songwriters, producers and artists aren’t proud of their work, what’s the point? Why do it in the first place?
Songwriting is spiritual. It’s mystical. The best songwriters tap into a higher power and channel a message through their art. Isn’t that the point of art? To reveal truths about life and the human experience? To help people feel understood and less alone?
Now, I’m not naive to think that every song is made this way. Most pop, country, R&B, and CCM songs have always been formulaic. There have always been brilliant leaders in these genres and then hoards of copycats.
There’s Chef Nobu and grocery store sushi. One meal you’ll remember for the rest of your life and the other is forgotten immediately after consumption.
And there’s a market for both. The innovative artists with perspectives build careers, fan bases and legacies. The rest is commoditized, forgettable, anonymous candy that is sweet to the taste, but empty at its core.
The major labels and publishers are in the business of hits at scale. So, when they see something that works, they try to mimic it a million times over. Instead of investing in art they believe in, they flood the market with uninspired clones to appeal to the lowest common denominator, hoping it’ll catch an algorithm. It’s trial and error. And since there are now no restrictions for the amount of music to release and test, no printing or distribution costs, the labels pump out as much music as they possibly can to manufacture a hit.
And when songwriters/producers are treated like factory workers instead of valued individuals with perspectives, they will surrender to get paid. Because at the end of the day, it’s all about who’s writing the checks.
How did we get here?
Well, for the past decade, Spotify has rewarded volume, consistency and “perfect fit content.” The most popular streaming platform in the world realized that their users were having choice paralysis when they opened the app and they wanted music to fit their moods and activities. Functional music, they called it. Music for studying. Music for meditating. Music for working out. Road trips. Jazz dinner parties. Chilling. Partying. You get the gist. And as Liz Pelly so expertly revealed in her book, Mood Machine, Spotify realized that their users didn’t necessarily care who the artists were behind the music, as long as the music fit their activity. Spotify found ways to pay a lower royalty rate for this background music to the companies providing it and then turned around and prioritized it in playlists and the algorithm.
+Spotify Commissioned Cheap “Fake Artists” And She’s Got the Proof
Functional music libraries licensing this type of music to Spotify at a discount initially hired studio musicians to make generic background music for popular Spotify playlists. Now, they don’t even need to hire musicians. They can type some prompts into Suno and pump out a constant stream of passable background music virtually for free. And the listener is none the wiser. And Spotify seems perfectly content with this arrangement. On the contrary, it seems Spotify actually promotes gen AI music as reported by virtually anyone with a Spotify account and the subsequent millions of monthly listeners publicly indicated on the Spotify “Artist” profiles. Not to mention these AI “Artists” who have actually charted and gotten multi-million dollar record deals.
A recent Deezer study revealed that 97% of people can’t tell the difference between fully AI-generated music and human made music.
That is how good the programs have gotten. Even though according to a recent Luminate study over 50% of Gen Alpha and Gen Z respondents said they have discomfort with gen AI music, and like music less once they realize it’s AI generated. But if they can’t tell, what’s the difference?
Every DSP is taking a different stance right now. TIDAL and Soundcloud have completely demonetized wholly generative AI music. Deezer detects and labels AI music and removes it from all of their editorial playlists. Spotify and Apple Music have requested that record labels, artists and distributors self-report (lol) and only then might label it. And Bandcamp has banned AI music outright.
DIY Distributors are starting to block distribution of songs they detect that have used generative AI. CD Baby president Molly Neuman told me on the pod that they aren’t interested in distributing gen AI music. Tunecore’s policy is to block music that has used “generative AI tools that rely on datasets that were not properly licensed” i.e. Suno. I appreciate distributors taking a stand and blocking gen AI music. But they can’t implement a policy without an appropriate and effective appeals process. Ukrainian artist Andriana-Yaroslava Saienko recently had her music blocked by Tunecore for distribution, claiming she used improperly licensed generative AI tools. She states she had not. Has Tunecore learned nothing from the “fraudulent stream” takedown debacle?
View this post on Instagram
At this stage, there’s no financial incentive for DSPs to label, demonetize or ban AI music. The listeners can’t tell the difference. And now the CEO of the Recording Academy – the organization behind the Grammys and, gasp, artist advocacy (?!?!?), is claiming that EVERY artist, songwriter and producer in pop, R&B and hip hop rooms are using it. Why would they change?
+Spotify Removed Your Music, Now What?
First off, let’s address WHY any producer, songwriter or Artist would use generative AI when they have the skills to make music themselves.
For one, as previously stated, Spotify rewards volume. The more music you put out, the more chances you have at getting picked up by the algorithm. No human can create at the pace machines can. And for the music being prioritized by Spotify’s official editorial playlists AND – more powerfully – algorithmic playlists, good enough is, well, good enough.
According to a Suno investor pitch deck from late 2025, Suno generated 7 million songs a day. Let that sink in.
And why would some producers and songwriters of “pop, R&B and some genres of hip hop” use generative AI? Well, even though they are not making music for background “functional” playlists, the incentive structure is still the same.
The way that streaming pays is on a pro-rata basis. The bigger the piece of the streaming pie you get, the more money you’ll make.
Every song competes against every other song. It doesn’t matter if fans listen to artists, it matters how much of the global recorded music pie you can muscle away from others. It is cut throat, capitalistic competition by nature. I get more, you get less.
+Why Spotify’s New Payment Model Falls Short For Emerging Artists
These producers and songwriters are being pressured by their labels to create more hits faster for cheaper. Because the labels are being pressured by the Spotify algorithm to deliver more music more frequently. So, instead of hiring a bassist to come in on Thursday, just ask Suno to generate a bass line right this moment. What would a dope horn section sound like on this track? Prompt it. Boom, Earth, Wind and Fire’s horn section in the house for just a $10 subscription.
And sure, I get that there is a ton of grey area here. Everyone uses plugins in their DAW. And most plugins are built on AI technology. But there’s a difference between typing a prompt for a full production and using Melodyne.
Some producers can’t sing, so they develop the instrumental production on their own and then have an AI voice sing a top line.
And for a demo that will not get released, I have no issue with this. But I would argue, make them do the work to find a human singer. If they don’t know any, use SoundBetter or go to a local open mic. Don’t have a budget to pay them? Fine, do a royalty split. There are millions of singers. Make them find human collaborators. Build community. Touch some grass.
+When It Comes To Royalty Splits, Be Fair, Not Generous
Or maybe a singer doesn’t have the expertise to make an instrumental production so they create one in Suno to sing over. Or a producer needs some funk guitar and no Splice samples are hitting. Or what about adding orchestration? What indie artist can afford to hire an orchestra? With a few prompts in the right program, it’s done. And yeah, songwriters are turning to Claude or ChatGPT the way they used to turn to rhyming dictionaries and thesauruses.
Personally, I have no issues with most of this. If songwriters, artists and producers want to use AI tools (ethically sourced tools – not ones like Suno or Udio that were built by stealing) as a tool for demos, fine. And frankly, I have no issue if a dad wants to make a goofy AI song for his daughter on her birthday to play for the party.
Where I draw the line is monetization. AI-generated songs should not pull royalties away from human made music.
If you want to enter the music industry by tapping into the royalty pool, then you gotta make music like a human.
Not to mention, the process of songwriting is human. It’s joyful, cathartic, challenging, exhausting, frustrating and mystical. Meditative, expansive, chaotic and jubilant. Why would you want to outsource this experience?
Deezer recently reported that 50% of all new music being delivered to its platform is wholly AI generated.
That is 90,000 fully AI-generated tracks every day. Since pretty much every distributor sends music to all DSPs, it’s safe to assume this holds true for Spotify and Apple Music as well. And Deezer also reported that 85% of AI streams are bot-driven, fraudulent streams.
So it’s clear, the vast majority of fully AI generated tracks being distributed to streaming services are for the intention of gaming the system to divert money from the royalty pool away from professional, human artists to these generative AI tracks via fraudulent methods.
So what should we do as an industry?
Well, for one, folks in the position of power and influence like Harvey Mason Jr., CEO of the Recording Academy, should not just say that everyone’s doing it so it must be ok, but should actually follow up that statement with: “…but we are drawing a hard line in the sand to advocate on behalf of human Artists making music with ethically (and legally) trained creation tools.”
We should work as an industry to pressure Spotify, Apple Music and the rest to label and demonetize wholly generative AI music.
Deezer has the tech to do it. Why can’t Spotify or Apple? They’ll claim they don’t want false positives. Sure, that makes sense. I get it. The last thing I’d want is my song that I wrote on an acoustic guitar with my heart, soul and higher power, and recorded live in studio with real human musicians, to be slapped with an AI label.
But Spotify saying “we don’t want false positives,” and then stopping there is lazy and disingenuous. But guess what? There’s an easy fix!
IF there is a false positive, then require the artist to provide proof it was human-generated. This ain’t hard for anyone who has ever made music. Provide stems. Provide a screen recording of the studio session. Show a video of the musicians or producers making the track. Show the singer singing it. Have a list of 20 different ways to verify their song was not generated by a prompt.
Require this verification upon ingestion. Previously, DSPs never required songwriter information. Now they do. They used to not display lyrics or credits, now they do. The DSPs should require proof of authentication upon ingestion. We can come up with this standard as an industry that every distributor and DSP adopts.
If AI music is demonetized, this problem goes away overnight.
No more incentive to game the system. And if we move to a user-centric payment model – whereby listeners’ subscription money only goes to the artists they actually listen to (instead of the global royalty pool distributed based on market share) – then most streaming fraud gets wiped out.
There are ways to solve these problems. But doing so requires more than just identifying the issues. It requires taking a stance and advocating for a better future for human artists for generations to come.
