DistroKid has started passing AI detection information to streaming platforms when an external system identifies a track as AI-generated.
According to DistroKid's own support documentation, the tag does not prevent a release from being distributed. It functions as metadata, with each streaming platform deciding how that information is ultimately displayed or used.
That makes this development less about banning AI music and more about classifying it.
We think that could ultimately make AI music easier to identify and potentially easier to discover. But it also introduces a harder problem for artists whose production processes do not fit neatly into either an AI or non-AI category.
ACRCloud is deciding what gets flagged
DistroKid says the detection comes from ACRCloud, an independent audio recognition company that provides technology for music identification, copyright compliance, broadcast monitoring and other large-scale audio analysis.
ACRCloud is not a new company built specifically around the current AI music boom. Its broader recognition infrastructure works with music services, distributors, labels and rights organizations, and the company says its music recognition database contains more than 150 million tracks.
AI detection is a newer addition to that infrastructure.
In January 2026, ACRCloud launched its AI Music Detector, initially positioning it for streaming platforms, distributors, collection societies and rights holders. The system analyses audio and returns a probability that it was AI-generated, while also attempting to identify the model responsible.
The detector currently supports music associated with systems including Suno, Udio, ElevenLabs, Mureka, Riffusion, Google Lyria and Boomy. ACRCloud has continued expanding that coverage as new models enter the market.
Importantly, the detector does more than produce one score for an entire song. ACRCloud says it can analyse individual segments, vocals and accompaniment separately.
That already hints at why AI music classification is becoming complicated.
You can test the same detector yourself
ACRCloud also provides a public version of its detector.
Its free AI music checker currently allows three checks per day without requiring an account. Users can upload a track and receive AI probability estimates for different sections, along with separate analysis of vocals and accompaniment.
The results should not be treated as proof.
ACRCloud's own technical documentation describes its results as probabilistic model outputs rather than definitive evidence of a track's origin. Its current system classifies audio as AI-generated when the calculated AI probability reaches 50 percent, while warning that real-world performance can vary after editing, compression, mixing and other transformations.
That caveat becomes particularly important now that the technology is being connected to distribution metadata.
A wrong label can still matter
DistroKid acknowledges that third-party AI detection is not perfect.
Artists who believe their music was incorrectly classified can challenge the AI tag through DistroKid. DistroKid says it will review the dispute, typically within 24 to 48 hours, and notify streaming platforms if the tag should be removed.
For now, DistroKid also stresses that the tag itself does not stop distribution. But classification can still matter even without an immediate penalty.
Listeners may interpret an AI label as meaning that most or all of a song was generated. An artist who used no generative AI could therefore face an obvious problem if their work is falsely flagged.
There is also a broader question about what happens if platforms increasingly use AI metadata for recommendation systems, filters or discovery tools.
A tag that begins as informational metadata does not necessarily remain informational forever.
The industry is moving toward AI disclosure
DistroKid's change is part of a wider shift across music platforms.
In October, YouTube Music introduced visible labels for tracks where generative AI played a significant role. YouTube says this includes cases where generative systems created major elements such as the composition and vocals.
Spotify has taken a more granular position.
When announcing support for AI disclosures in music credits, Spotify explicitly argued that AI use is “a spectrum, not a binary”. Its approach is built around an emerging DDEX industry standard that can disclose where AI was involved, including vocals, instrumentation and post-production.
This is important because detecting that AI appears somewhere in a recording is not necessarily the same as understanding how the song was made.
A song can be partly generated
We recently explored this problem in AI Music Is a Spectrum. Our Labels Should Reflect That.
Our argument was that music production increasingly sits somewhere between fully AI-generated and fully human-created.
Imagine an artist writes the lyrics, sings the vocals and performs every instrument except for one hook generated using an AI model.
Is that an AI-generated song?
Now reverse the situation.
An artist writes every lyric and melody but uses AI for the vocals and instrumentation. Most of what the listener hears may be generated, while much of the underlying authorship still comes from a person.
Those two tracks involve AI in fundamentally different ways.
Yet a binary AI label can make them look surprisingly similar.
ACRCloud itself is already confronting this problem
Interestingly, the technology behind DistroKid's tags is becoming more nuanced than the label itself.
When ACRCloud introduced segment-level detection, the company explicitly acknowledged that AI is often used for particular parts of a song rather than the entire recording.
Its system can therefore attempt to identify which sections appear generated and which appear human-created.
It can also analyse vocals separately from accompaniment.
A generated singer over a human-produced instrumental is not technically the same production process as a fully generated track. ACRCloud's own detector is increasingly capable of seeing that difference.
The question is whether the metadata eventually presented to listeners will preserve it.
Better AI detection could actually help AI artists
There is also a positive side to this infrastructure. Reliable AI metadata could eventually make AI music more discoverable.
Listeners interested in generative music could potentially filter for it. Platforms could create dedicated discovery experiences, albeit the demand for this might be somewhat niche. Artists who openly work with AI would no longer depend entirely on biographies or social media posts to explain how their music was created.
Better disclosure could also reduce some of the suspicion surrounding AI music.
Instead of asking whether an artist is secretly using AI, platforms could provide clearer information about where AI participated in the production process. The problem comes when detection is more sophisticated than disclosure.
AI detection is getting easier. Classification is not
We broadly support greater transparency around AI music.
Artists should be able to explain when they use generative systems, and listeners should have useful information about how the music they hear was produced.
But a single AI tag can only communicate so much.
It cannot tell you whether AI wrote the lyrics, created one bassline, generated the singer, supplied the entire instrumental or produced almost everything in the recording.
DistroKid's adoption of ACRCloud therefore looks like an important step towards making AI identification part of ordinary music distribution.
It also demonstrates why the industry will probably need something more nuanced than identification alone.
As we argued in our AI Music Spectrum, the increasingly useful questions are not simply whether AI was used.
They are what AI did, where it entered the production process and how much of the finished work it actually shaped.
AI detection technology is becoming increasingly capable of answering the first part.
The music industry still has to decide how to communicate the rest.




