Australia's ARIA Charts Ban Wholly AI-Generated Songs After Madonna Cover Controversy

Australia has become the first country whose official music charts formally exclude songs created mostly or entirely by artificial intelligence. On August 25, the Australian Recording Industry Association (ARIA) confirmed that wholly AI-generated recordings will no longer be eligible for the ARIA Charts, with the new rules taking effect from Friday, August 28. The trigger was a months-long controversy over a cover of Madonna’s 1989 classic “Like a Prayer” that conquered Australian radio before anyone knew how it was made.

What happened

The story centers on Josh Fawaz, a Brisbane DJ and producer who released an AI-driven rework of “Like a Prayer” in April 2026. The track climbed steadily — by mid-year it had reportedly become one of the most-played songs on Australian commercial radio and spent roughly 16 weeks in the ARIA top 20, topping the ARIA dance singles chart along the way. Listeners loved it. The problem was what they didn’t know: the vocals and drums were AI-generated, and the track carried no disclosure of that fact.

An investigation by Australia’s ABC, followed by public allegations from other producers, forced the issue into the open in July. Fawaz subsequently added explicit AI credits to the track on Spotify — but by then the damage to trust was done. Radio programmers admitted they had been playing a record whose “artist” was substantially a machine, and industry figures began asking an uncomfortable question: if nobody can tell the difference, what exactly is a chart supposed to be measuring?

The new rules

ARIA’s response, confirmed August 25 and reported by Reuters, ABC, the BBC, and Billboard, establishes three core eligibility conditions:

1. Recordings must be “substantially human-made.” A human performance has to carry the record — the vocal, the playing, the core creative act.
2. Generative AI is permitted only in a “supporting role.” Songwriters, producers, and artists can still use AI for tasks like mastering, stem cleanup, or arrangement experiments, provided that use doesn’t breach copyright law.
3. The recording must raise “no stream or chart manipulation concerns.” Ineligible tracks can be refused a chart survey or removed retroactively — meaning the rules carry enforcement teeth, not just a policy statement.

Fawaz’s “Like a Prayer” cover is widely expected to be the first casualty, with reports indicating it faced removal once the rules took effect.

Why this matters beyond Australia

This is the first time a national chart body has drawn a formal line between human-made and machine-made music, and the precedent lands amid a global scramble over the same question.

Streaming platforms are moving in parallel. Spotify’s industry-facing CONNECT platform has begun identifying and flagging AI-generated songs and artists, with visible “AI Generated” labels expected to roll out more broadly. Electronic music stores like Beatport have gone further, currently rejecting AI-generated releases outright. Meanwhile the major labels are trying to monetize rather than ban: Universal Music Group signed a deal in May allowing fans to create licensed AI covers and remixes of participating artists’ songs, with the artists taking a cut of revenue.

These are three very different answers — exclude, label, or license — to one shared problem: the marginal cost of producing plausible music has collapsed to nearly zero, and every ranking, royalty system, and recommendation feed built on scarcity assumptions is now strained.

The hard part: drawing the line

The obvious difficulty is enforcement. Detection tooling is reasonably reliable for audio that is fully synthesized — generated waveforms carry statistical fingerprints that differ from microphone recordings. But the “supporting role” boundary is far murkier. If a singer records a real vocal and an AI model replaces their voice, generates the drums, and rebuilds the instrumental, is that “substantially human-made”? What if AI wrote only the lyrics? Musicians and managers have been quick to point out that there is no settled standard, and ARIA has so far declined to publish a technical test — leaving the judgment in practice to disclosures, metadata, and industry scrutiny.

There is also an uncomfortable epistemic wrinkle: the Fawaz track succeeded precisely because listeners and programmers couldn’t tell. A rule that depends on provenance rather than sound quality implicitly concedes that the product is competitive — and shifts the fight from “is it good?” to “who gets to count?”

What to watch

• Whether other chart compilers — the UK’s OCC, Billboard, Germany’s GfK — follow with their own eligibility rules, and whether they adopt ARIA’s “substantially human-made” language or something stricter.
• Whether the first retroactive chart removal triggers legal challenge from a distributor or artist, which would force a court to define “substantially human-made” more precisely.
• How quickly AI-content labeling spreads from chart bodies and industry platforms to the consumer-facing surfaces where most people actually encounter music.

For now, Australia has drawn the industry’s clearest line in the sand: charts are for humans. Whether that line survives contact with technology that keeps getting harder to detect is the question every other music market is now watching.