The Australian Recording Industry Association ruled that wholly AI-generated tracks are ineligible for the ARIA Charts, effective from the chart dated Monday, 31 August 2026. Recordings that use generative AI in a supporting role remain eligible. It is one of the first national chart bodies to convert the music industry's general disquiet about generative audio into an enforceable eligibility rule with defined penalties.
Key facts
- ARIA announced the change on 25 August 2026; it takes effect from the chart dated 31 August, published 28 August.
- Under the updated Code, an AI-assisted recording is eligible only where it "is substantially human made" and "raises no stream or chart manipulation concerns."
- ARIA applies the definitions from the labelling standard announced by the global music community on 10 July, and implements principles set by the international body IFPI.
- Primary source: ARIA's announcement.
The distinction ARIA is drawing is the one the whole argument turns on. Producers have used machine tools for decades -- pitch correction, generated drum parts, stem separation, mastering assistants -- and a rule that treated any AI involvement as disqualifying would delete a large slice of contemporary music. So the test is not whether AI touched the recording but whether a human made it. Supporting role, eligible. Generated wholesale, not.
ARIA CEO Annabelle Herd put the reasoning bluntly. "Artists already use AI tools in their work, the Charts can and should evolve to keep room for that, but music generated wholesale by services built on artists' recordings is a different matter," she said. She added: "The ARIA Charts will always remain a transparent measurement of the music Australia consumes, but a chart that rewards unlicensed AI output would undercut the very basis of the recorded music we exist to represent."
That second sentence is the actual argument, and it is narrower and stronger than a general objection to synthetic music. The complaint is not that the output is machine-made; it is that the machines were trained on the catalogue the chart exists to measure. A chart that ranks a generated track above the recordings it was trained on is measuring a loop.
The enforcement provisions have teeth, which is what separates this from a position statement. ARIA can decline to accept a recording for survey, exclude or remove it from the charts prospectively or retrospectively, adjust chart positions, withdraw accreditations, and revoke or request the return of an ARIA number one award. An ineligible recording also cannot be nominated for an ARIA Award. Retrospective removal is the significant one: a track can chart, be celebrated, and then be unwound.
The obvious hard question is detection, and ARIA's release does not claim to have solved it. There is no described technical detector. Eligibility rests on the labelling definitions agreed by the global music community in July and on ARIA's own judgement about whether a recording is substantially human made -- which is to say, on disclosure plus adjudication rather than analysis. The Code adds a disputes process so artists can contest an exclusion, which is a tacit acknowledgement that these calls will be contested. Anyone following the content provenance and watermarking debate will recognise the gap between a rule and a way to verify it, and the same tension runs through platform-level labelling like Amazon's AI-generated people disclosures.
Why it matters: charts are not just scoreboards, they are the allocation mechanism for radio play, playlist placement and touring economics. Deciding what counts is deciding where money goes. Herd's closing line makes the ambition explicit -- ARIA called on "all parties who have a role in deciding the music played and promoted to Australian audiences, particularly radio, to support human artistry and implement similar changes across their own codes." This is a national body trying to set a template, and other chart authorities now have a working one to copy or reject.
The timing is not incidental. Generated tracks have been appearing on streaming platforms in volume for over a year, and several have charted in smaller territories, usually surfacing through playlist placement rather than an audience that sought them out. A chart is a survey of consumption, and consumption is measured through the same platforms where generated material is cheapest to flood. That is the manipulation half of ARIA's two-part test doing real work: a rule about human authorship is also, in practice, a rule about who can afford to produce ten thousand tracks a month.
The unresolved question is what happens to the middle of the distribution. A vocal delivered by a synthetic voice over a human-written song, or a human vocal over a fully generated arrangement, is neither wholly generated nor comfortably "supporting role," and those records exist in commercial quantity today. ARIA's answer is procedural rather than technical -- apply the July labelling definitions, judge whether the recording is substantially human made, and let the disputes process handle the arguments. That will work exactly as well as the labelling standard's definitions turn out to be precise, which nobody yet knows.
Originally published on Ground Truth, where every claim is checked against the primary source.
