ARIA Bars Fully AI-Made Tracks—but Assisted Songs Can Still Chart

The Australian Recording Industry Association has revised its chart rules to exclude recordings generated wholly or primarily by artificial intelligence while retaining eligibility for substantially human-made, AI-assisted music. The ARIA Charts Code dated 24 August 2026 makes the source of a recording’s central creative elements—not the mere use of an AI tool—the decisive test.

The change applies in Australia from the ARIA Chart dated Monday, 31 August 2026, published Friday, 28 August, according to ARIA’s implementation notice. Eligible recordings involving generative AI must also raise no stream or chart-manipulation concerns.

The change is a chart gate, not a ban on AI tools

The new boundary concerns eligibility for ARIA’s rankings. It does not prevent an AI-generated recording from being distributed, streamed or played on radio, and it does not determine royalties. Those decisions remain with the services and organisations responsible for distribution, programming and payments.

Supporting uses of generative AI remain possible. SBS’s account of the change states that AI-assisted songs can still chart when they are substantially human-made and do not raise manipulation concerns.

The narrower rule matters because an AI tool can appear at very different stages of production. A person using automated mastering is not treated the same way as a release whose lead vocal or principal instrumental performance was generated by a model.

How the substantial-human-contribution test works

The practical test begins with the source of the song and its central performances. ARIA’s chart-change FAQ divides recordings into AI-generated, AI-assisted and production-only uses:

1. Was the recording or its primary creative material generated? Entirely prompt-generated music is ineligible. The same applies when generative AI supplied a lead vocal or a key instrumental performance, even if a person later mixed, mastered or edited the result.
2. Did people create and perform the core work? A track remains potentially eligible when people wrote the song, performed the lead vocal and played the primary instruments. Minor generative additions can then be assessed as assistance rather than replacement.
3. What role did the generative element play? AI backing vocals, a limited non-primary instrumental part or a transformation derived from a human performance may fit the assisted category. A voice-to-voice translation based on a human performance is one example identified in the FAQ.
4. Was AI confined to production? Automated mastering, stem separation, drum-machine functions, short background samples, reverb and instrument patches played by a person are treated as negligible or non-material uses. They do not by themselves make a recording AI-generated.
5. Does the recording clear the other integrity rules? Substantial human contribution is necessary but not sufficient. The release must comply with applicable law, including copyright and related rights, and must not present stream or chart-manipulation concerns.

The framework sets no percentage of required human input and no fixed formula based on credits. ARIA assesses each recording on its facts, which leaves discretion where human and generated performances are difficult to separate.

Using a lawful or authorised AI service does not resolve the creative test by itself. The recording can still be ineligible if that service generated its primary creative elements; conversely, a minor AI contribution does not remove the separate obligation to comply with copyright and other applicable laws.

An eligibility decision can reach an existing release

Every release submitted for an ARIA survey now requires an accurate declaration about generative-AI use. If a credible concern arises, ARIA can seek more information from the rights holder and allow the artist or representative to provide evidence about songwriting and the central performances before making a decision.

An adverse decision can keep a recording out of a survey or remove it from the charts prospectively or retrospectively. ARIA can also adjust chart positions, withdraw accreditations, revoke a number-one chart award and make the recording ineligible for an ARIA Award.

A disputed assessment can be referred to the ARIA Chart & Marketing Committee. Its recommendation then goes to the ARIA Board, whose determination is final under the Code; disputes and decisions remain confidential unless ARIA authorises disclosure.

The rules also distinguish versions and album tracks. Replacing generated central elements with human performances creates a new recording that can be assessed independently, without inheriting the earlier version’s chart history. Albums are considered track by track, so an ineligible recording does not contribute to the release’s chart calculation.

The rule follows scrutiny of an AI-assisted chart release

The immediate context includes Queensland producer Josh Fawaz’s dance cover of Madonna’s “Like a Prayer.” ABC’s coverage published on 25 August 2026 linked the policy debate to the track, which had reached number four on two ARIA charts in July and later carried credits identifying AI-generated vocals and drums.

That history illustrates the boundary but does not establish a public ruling on the track under the revised Code. ARIA does not plan to publish a running register of assessed or excluded recordings, so rights holders will generally receive decisions directly.

The first affected chart will show the visible effect of the new eligibility gate, but not the reasoning behind every decision. The unresolved issue is how consistently ARIA’s case-by-case standard will handle recordings in which human performance and generative transformation are closely intertwined; the association intends to keep the Code under review as the technology develops.

Also read:

• ‘Made by Human’ — The New Trend in Content Labeling, and Why It’s Raising More Questions Than Answers
• AI Agents Work Longer, but Research Still Ends at a Human Gate
• 44% of All New Music on Deezer Is AI-Generated — But Almost Nobody Is Actually Listening to It