SH Labs Data Reveals Over One Million Songs with 40% Showing Signs of AI

More than a million releases were analyzed by SH Labs, SubmitHub’s AI Music Detector, in July alone, producing a large real-world dataset on AI-generated music. The results paint a striking picture of how rapidly artificial intelligence is reshaping music creation: 23.2 percent of analyzed recordings were classified as fully AI-generated, while an additional 15.3 percent showed evidence of AI-generated audio that had subsequently been modified or processed by humans. Together, nearly 40 percent of all analyzed tracks showed signs of AI involvement, underscoring the growing prevalence of hybrid creative workflows across today’s music industry.

SubmitHub’s data also points to a growing disclosure problem. In 2026, 31 percent of artists whose music was identified as containing AI-generated audio explicitly indicated that they had not used AI. Because SubmitHub asks artists directly about AI usage during the submission process, the discrepancy offers a rare comparison between self-reported AI use and automated detection at scale.

The finding suggests that disclosure requirements alone may not be sufficient for platforms, distributors, and rights organizations seeking to understand how AI is being used in music.

SubmitHub founder Jason Grishkoff states:

“I think the path forward for AI music is disclosure. People should be able to decide for themselves whether they want to engage with AI-generated music. Our job is to give them enough information to make that choice.”

Behind the numbers is a platform that has quietly become one of the music industry’s most widely deployed AI detection systems. Originally developed to help SubmitHub’s network of curators filter submissions according to their individual preferences, the technology is now used by organizations including Bandcamp, GEMA’s MusicHub, Traxsource, SlopTracker, numerous distributors, digital platforms, and one of the three major labels. Having surpassed one million detections in July, the platform is already on pace to process more than 1.5 million analyses during August.

Unlike many AI detection systems, SubmitHub’s Detector was built to recognize the messy middle between fully human and fully AI-generated music. Because SubmitHub encounters recordings across the full spectrum of AI use, the system is able to distinguish between fully AI-generated works and recordings where AI-generated material has been substantially modified, reworked, or combined with human-created elements. That hybrid category accounted for more than 15 percent of all July analyses.

That distinction is becoming increasingly important as music companies develop policies around AI-generated content. A track created entirely by an AI model may raise very different questions from one where AI-generated material has been incorporated into a substantially human-produced or human-edited recording. SubmitHub’s Detector gives platforms, distributors, rights organizations, and curators a more nuanced signal than a simple AI-or-human flag, allowing them to apply their own policies and preferences accordingly.

Performance has also driven adoption. On a confidential industry benchmark, SubmitHub’s Detector achieved greater than 99 percent accuracy, results that helped prompt one of the world’s largest music companies to begin integrating the technology into its own workflows. Its ability to identify both fully generated and hybrid recordings has emerged as one of the system’s defining strengths as AI becomes increasingly embedded in conventional music production.

As AI becomes a routine part of music production, the challenge is shifting from determining whether it exists to understanding where it was used, how extensively it was used, and whether that use was disclosed. With more than a million analyses completed each month, SubmitHub’s AI Music Detector is helping the industry answer those questions at scale.