The scale of AI's presence in recorded music is becoming measurable, even if the numbers come with a sizable margin of dispute. SubmitHub, an online platform for playlist and blog submissions, analyzed over a million tracks using its detection tool SH Labs and found that 23.2 percent were fully AI-generated. An additional 15.3 percent contained AI-generated audio that had been modified or processed by humans. Together, that suggests nearly four in ten recent releases involved AI at some level.

SH Labs, which is also used by Bandcamp and Traxsource, was introduced in June as a response to the rising volume of AI-assisted submissions. The platform sees detection as a disclosure mechanism more than a gatekeeping tool. Founder Jason Grishkoff frames the approach directly: "People should be able to decide for themselves whether they want to engage with AI-generated music."

But the detection layer is far from settled. SubmitHub reports that 31 percent of artists flagged for AI use this year denied using any AI tools when contacted. That opens two possibilities: either the detector is over-reaching, or a meaningful portion of creators are not disclosing their methods. Both likely contribute.

The tools exist, adoption is widespread, and the industry's vocabulary for admitting who used what is still forming. For now, the numbers are less a verdict than a snapshot of an ecosystem adjusting to its own automation.