You might assume that some of the best-selling songs in music history have their paperwork in order. After all, recordings by Michael Jackson, Fleetwood Mac and the Eagles have generated decades of revenue, backed by some of the world's biggest publishers.
But a new white paper suggests that even music's most valuable catalogues contain surprising metadata problems, including recordings connected to entirely different songs.
The study from AIMi Labs, titled Who's Eating Your Lunch?, used an AI-powered music-rights research tool to examine publicly available data from the Mechanical Licensing Collective (The MLC).
Researchers examined 111 tracks across ten of the best-selling albums of all time, including "Thriller," "Rumours," and "Hotel California." All 111 tracks were registered, with average recorded ownership-share coverage of 99.3%. Yet 56 tracks (50.5%) were associated with multiple musical-work records requiring further investigation.
That doesn't mean half the recordings contained errors. Remixes, adaptations and alternative versions can legitimately involve multiple registrations. But individual investigations uncovered some remarkable inconsistencies.
When one hit becomes another
Fleetwood Mac's "Don't Stop," for example, was linked not only to Christine McVie's composition but also to Queen's "Don't Stop Me Now" and Journey's "Don't Stop Believin'."
Recordings of Fleetwood Mac's "Dreams" were associated with eight additional work records beyond Stevie Nicks' correctly registered composition. Meanwhile, Michael Jackson's "The Lady in My Life" was connected to "Thriller," another composition written by Rod Temperton.
These aren't simply misspelled titles or missing credits. They're examples of recordings associated with the wrong underlying musical works, potentially affecting how composition royalties are allocated. Pete Whelan, founder of Program Music and AIMi says:
"Being matched to a song is not the same as being matched to the right song. A registration can look complete and still point to somebody else's song."
The study cannot determine whether these connections resulted in incorrect payments or how much money might be involved. The MLC's public data doesn't disclose individual royalty transactions.
What about everyone else?
If these inconsistencies exist in some of the world's most commercially successful catalogues, the implications for smaller artists are worth considering. Independent songwriters and self-releasing musicians may lack the publishing teams and resources available to major artists. The paper identifies completely missing registrations as a particular concern for creators without professional publishing administration.
But the larger finding goes beyond missing information. A song can be registered, its ownership shares can appear complete, and its recording can still be connected to the wrong composition. That's a problem that simply improving registration rates won't necessarily solve.
Could AI help fix the problem?
There's an interesting irony to the research. While AI remains a contentious issue for music creators, AIMi demonstrates how the technology might also help protect their earnings. The platform has investigated tens of thousands of musical works, identifying contradictions in rights metadata and preparing evidence for human review.
A free catalogue-auditing service for songwriters is scheduled to launch at Who's Eating Your Lunch? by the end of 2026.
The research also makes a broader argument for transparency. These discoveries were possible because The MLC made its data publicly accessible. AIMi is calling on other collection societies to provide similar opportunities for independent scrutiny.
