These days roughly 20% of the songs coming through our platform for promotion are AI-generated. Roughly 75% of them are honest and declare their AI usage - but another 25% try to hide it. Some of them are actually writing scripts to "clean" their audio so that it can bypass detection.
Do not try to solve an unsolvable problem, you'll end up hurting real users quite a bit more than you might imagine. Imagine new enthusiastic users trying your platform getting hit with an AI label because of inevitable false positives.
'Detecting AI' is not a problem that has real solutions, the only avenue is something supply side like synthid. But that harms users too, by introducing further barriers for indie users.
I train music generation models. They are very trivial to detect. In fact, detecting them then training them to evade detection by the detection model is a big part of training them! But the detectors win instantly without some hardcore regularization. Simply turn that off and you've instantly got a perfect classifier.
This isn't like text classification, the signal many orders of magnitude higher bitrate and so many more corners need to be cut. It's likely going to be nearly impossible or at least not remotely worth it to generate an audio signal that is truly undetectable in the foreseeable future.
You are right, the output of a model that generates music directly is, for now, easy to categorize as AI.
What this big flux of AI generated music online isn't really that. It'a a tiny bit autogenerated stuff and a whole lot of automatically remixed stuff. The reason it can not be easily classified as AI is because quite a bit of human produced music is also that, and you'd just shut out real users.
Today. Trying to detect AI is like extracting water from puddles in a lake that is quickly drying up. What is the point in the short term if it's impractical in the long term? It will catch some low-hanging fruit in the best case, and will find false positives in the worst.
My point is you should consider creating truly undetectable audio end to end with AI to be effectively impossible for the foreseeable future (i.e., I would bet money it is still trivially detectable five years from now). It won't be detectable to humans, though, only models.
in the broad strokes of ai generated, i wouldnt be so sure.
if the ai picked a bunch of samples and combined them together and mastered using an mcp to a DAW, how is that particularly distinguishable vs a person doing the same thing badly?
i can see how the llm generation pictures of spectrograms is essy to spot, but much less so with tool following.
even worse of you using a vla to have it actually play the guitar and use the recording as a sample.
theres some time and setup to make it happen sure, but somebody put that all in a studio and expose an mcp
This is an aside, but thank you for doing this work! As a musician who plays real instruments and submits real songs to Submithub, it's nice to know that hard work is going into validation and prevention of scammers passing off AI as their own talent. Keep fighting the good fight.
"AI detectors" are fun like horoscopes are fun, until they flag your music as AI generated, and distribution channels blacklist you and your label sues you. On the bright side, you can sue the creator of the AI detector in return.
I've had my digital art flagged a few times for various reasons (automatic copyright infringement and NSFW filters) -- so this is nothing new (in particular the artwork blocked the upload for some artist songs). The only thing is to have a reasonable appeal process. In all cases we got an automated approval after appeal, but it can put an untimely delay.
Honestly I hope that the AI filter would be much better in terms of false positive than the aforementioned one, if only because it should be easier via statistical methods.
The only reason you're saying that is because you haven't tried to build such a detector yourself. It's not like text where it's impossible to tell reliably if something's AI generated or not, from a technical perspective it's very trivial to detect anything coming straight out of a Suno/Udio prompt.
Nobody open sourced their detection algorithm as that would just trigger a cat-and-mouse game between Suno/Udio and a detection platform (and Suno/Udio have way more VC money than you do), but plenty are being sold as a service and work very reliably.
> …from a technical perspective it's very trivial to detect anything coming straight out of a Suno/Udio prompt.
It's trivial to vibe-code something that detects watermarked output and accidental model fingerprints. But next week the watermarks will be defeated, and the accidental fingerprints will change and ultimately disappear. It's not possible to generally solve the "To what degree is this audio AI generated?" problem, any more than it has been to solve the same problem for text and images. https://mitsloanedtech.mit.edu/ai/teach/ai-detectors-dont-wo...
You're discussing pure hypotheticals, I'm discussing what you can do today with very little effort. If that changes, it changes, but so far it's trivially easy.
The question I'm more interested in is why other music streaming services are not interested in doing this trivially easy work to get rid of spam, even if it's just for the short run as you assume it will be.
Has this happened to you? Or anyone you know? Or do you know of a lawsuit by a label against an artist for making AI music, and a lawsuit by the same artist against an AI detector for flagging a false positive. This story seems extremely unplausible.
Aside, your analogy doesn’t make sense. Horoscopes are generally not in the business of signal detection, and are usually enjoyed by the reader of the horoscope, like any other art. If you had used a sudoku solver your analogy would make a bit more sense.
do you have any idea on what percentage of musicians use AI to create the song and then also create the sheet music so they can play it themselves? That seems like a decent workflow, use AI to get the song right, and then record yourself playing it with you're own creative tweaks. That's kind of how I do AI assisted coding.
I assume you're not a musician, because that sounds insane. If you're good enough to play at full speed from brand new sheet music, then you don't need the AI. Playing from sheet music isn't like typing.
There are some composers who use a workflow like this - Suno is a scratchpad which can be used to quickly trial ideas, clarify concepts with collaborators, etc. don't think it's common, either among composers, or Suno users at large
I'm curious how your platform might avoid false positives with intentionally repetitive music, in particular techno (either produced via a DAW or hardware).
Our relationship to slop is a bit more complicated than that, no?
Whether it's terrible music or not is somewhat irrelevant. Plagiarized music doesn't sound worse than the original.
What seems to matter more is the story behind the work. Basically, if the author is a grifter trying to make a quick buck, then it's slop. You can make an argument that Taylor Swift qualifies as slop, but most people will disagree. The public will be the final arbiter. All I really want is a big red lever to cast my vote.
Great to see this! P&P is one of the OG music blogs.
For context, I started music blogging in 2007 (my site is still around - Indie Shuffle - indieshuffle.com). I also started SubmitHub (submithub.com).
About this: "we’re not sure what a music website’s role is in 2026 and beyond".
I waffle between cynicism and optimism. Music blogs used to be major influencers. The rise of social media and streaming platforms squashed a lot of that, and our audiences have dwindled.
Meanwhile Spotify is increasingly pushing toward AI recommendations rather than human curation. I've heard rumor that their editorial team has been halved.
So, where do music blogs fit in? Will there be a resurgence in their audience? The cynic in me says "no". In general, blogs have gone out fashion, and users don't seem to have the patience to listen through a mountain of unknown music.
But there are still those diehard music lovers who do sift through the hundreds of thousands of daily new songs. And there needs to be a human touch to curation somewhere along the line - a space that blogs still fill.
I suppose at the end of the day I'm mostly just blogging for me. For the artists I share, there's some minor exposure - as well as SEO and AI ingestion. I don't think I can make or break an artist anymore - not like we used to back in 2010. But my blog is an extension of me, and I hope that for Jacob there's some similar upside.
I think about this all the time through the lens of "authority" on a topic. When we yielded our gathering spaces online to major social sites (read, Zuck et al) we then gave the content all the authority of what it means to dialogue in those places. Which is to say... not much.
This has impacted journalism, music, science, and so much more. It would take an eternity to hash out my perspective but I think that there's value in realizing that. And I think there's value of creating content from the authority of a personal website with cache. I think music is a great place for this to take off, since you don't need institutional backing. You just need good words and a deep connection to the community. In that way, I hope people do write and create good content through their own mediums/sites. And I hope we all join in reading and sharing those sites.
Thank you for creating/maintaining indieshuffle! I visited 12-15 years ago with very fond memories. Your site was one of the few I checked on rotation due to good curation. Thank you for the human hand you played in music discovery! Visiting the site again today and “it feels very similar to how it was back then”. Thank you for sharing such extension of yourself!
I’m not a musician, but I have referred several friends to SH! You’ve helped them a ton, thanks!
Re the roll of blogs, I’m guessing that to an extent at least Spotify and other recommendation algorithms/bots are taking into account the buzz online and what bloggers are saying?
I've been working on an AI detector for the last few months. Updated it to handle Suno V5 on Wednesday - looks like it's very similar to V4.5. Am curious to see how this Studio version impacts the model I've trained.
After login it's free. But my site has been targeted by a lot of spam/abuse over the last decade, and login is something I've needed to set up to avoid that :(
I make my living off these platforms (primarily the second). So in essence, my discovery-centric services are viable products. That said, I'm not sure that's 100% what he was after in the Twitter thread this article was based on: https://twitter.com/jherskowitz/status/1466078600822677513
I’m a musician and Indie Shuffle was my first “break.”
It gave about 20,000 plays which BLEW MY MIND at the time. Nothing like waking up to a huge increase
Today I’m a modest success. Several songs have 1-2m plays on Spotify and I make $800 / month from streaming. It’s just something I do in my evenings for fun.
I owe my success to outlets Indie Shuffle and SubmitHub—- I’ve found Spotify really privileges discovery for major label artists.
I appreciate the counter-take — it seems like almost every take I read on the modern music business is coming from people who don't actually know what the reality on the ground is.
And gd submithub is awesome, I have been sucked right in, making submissions, buying credits, rating songs. It's taken up my whole morning, well done!
Sounds like a perfectly fine mismatch between proper bootstrapping and the mindset of growing investment fueled by some hypothetical value proposition.
Is submithub what I think I am seeing? Basically a solution to a spam problem by offering a channel that requires the equivalent of stamps so that senders rate-limit themselves, focusing a bit more on quality over quantity? If that's not a complete misperception I like it very much, great niche-spotting!
What I thought "Discovery" was is finding new music given some other music preferences - like Spotify's curated playlists or Song Radios. Submithub doesn't fit that to me - it's more like a social network (and I guess you make your money the same way, via advertising).
As someone who pulled this off, it was less about discovery and more about being part of it already. The problems are much easier to understand and address when they're actually problems you're facing yourself.
Mine isn't anywhere near paying the bill yet, but it's at 60+ customers for $20/year.
It's exactly as you described. I had a pain point which was trivial for most, but not for me. No existing solutions solved it properly so I decided to make it myself.
While searching for a solution, I found that a lot of other people had the same problem. It was bad enough for them that they took time out of their day to write their complaints online.
In solving my own problem, I was creating something that other people might've pay for. And they did!
I quickly learned that marketing was the most important part of the process.
Hey, two quick suggestions, mostly related to the actual beatoftheday.org website:
1) Put some info on your homepage about what this actually does. When I first landed I wasn't sure what the purpose of the website was.
2) Resize your images! For example your header logo is massive and took a good 5+ seconds to download. You have a max-width of 994px, but the image itself is a whopping 14,976px wide!
Some info would go a long way. I'm still not sure what kind of collaboration you are referring to. Are we talking collecting mp3 files? Curating playlists? Or creating music?
There's a link tucked away in the upper right (on desktop) labeled "Info" that explains the site. I agree with you though that at least a sentence or two on the main page would be smart.
I'd like to be able to sort by 'liked' and by 'baked'. If I'm looking for a new track to collaborate on I want to see the ones with the fewest 'baked' ratings.
Also, I'd love for there to be a way to see the full lineage of a given track. Perhaps even be able to fork it if, e.g., I liked what contributors 1, 2 and 3 did, but not #4, I could work off of #3's rendition.
SubmitHub is not paying to get added to playlists. It's paying to guarantee that your song is considered by the playlister - with no promises that it's actually shared. If they do decide to share it, no additional money is required.
You're basically saying: "Hey, if you give me 3 minutes of your time to listen to my song and let me know what you think I'll give you $1."
For reference, roughly 1 in 5 submissions end up getting shared. So if you send to 100 playlisters, 80 of them will say "thanks but no thanks" and roughly 20 of them will add it to their playlist for no additional cost.
Hey, SubmitHub founder here. Sorry for the confusion. You get two types of credits: standard (the 2 you saw) and premium (which you have to buy). Your standard credits refresh every 4 hours (assuming you use them).
When going through the submission process you'll see it prompts you whether you want to use your standard (free) credits or premium credits. If you don't have any premium ones -- and don't want to buy them -- make sure you stick to the 'standard' path.
The core idea behind SubmitHub's "model" is that for decades it's been near-impossible to catch the attention of bloggers/playlisters/curators/whatever. Our system dangles a carrot (~$1) in front of them to guarantee a response+feedback about your song (with the ideal outcome of course being that they share your song - no additional costs involved).
And I appreciate you are matching musicians with audiences, super valuable service. This particular credits ambiguity registered as a pretty grey pattern. My expectation was just say "$10 to put this in front of an audience of x," or "share a song with our network and then buy premium features," instead of breadcrumbing the incremental commitment screens before the hard upsell to complete it. I'm sure your mixpanel or other metrics show your abandoned carts data and if this pattern is working, it's working.
For pro musicians who are committed to your product as a promotion path, perhaps they don't care about that credits ambiguity because they go in with a marketing budget and know what they're going in for. For me, I will stick with making music for fun. I wish your team success at their mission.
Hey, SubmitHub founder here. Top tip is to do a bit of research before jumping into it - it will dramatically improve what you get out of it. Here's a really good recent podcast by an artist who managed a 68% approval rate: https://pod.co/bandhive/submithub-success-strategies-steve-m...
These days roughly 20% of the songs coming through our platform for promotion are AI-generated. Roughly 75% of them are honest and declare their AI usage - but another 25% try to hide it. Some of them are actually writing scripts to "clean" their audio so that it can bypass detection.