A song made with an AI tool has now charted on the biggest singles chart in the world, and almost nobody noticed until months later. That single fact has forced an awkward question into the open: is AI music actually allowed on the Billboard charts, and who decides?
The short answer is that no chart has a binding rule yet. The longer answer involves eleven record companies, a detection system that listens for artefacts human ears cannot hear, and a survey suggesting almost none of us can tell the difference anyway.
What actually happened with “Rubberz”
In 2026, a synth-pop track called “Rubberz,” credited to Los Angeles rapper Fenix Flexin, climbed to No. 58 on the Billboard Hot 100. It sounded nothing like his previous work — an AI-rendered vocal singing in a British accent, closer to 1980s new wave than West Coast rap. Listeners immediately accused the track of being machine-made. The artist and his producer denied it for months.
On 7 August 2026, Fenix Flexin reversed course and acknowledged using the AI music platform Treblo, arguing the tool had not changed the recording process itself. His public defence was blunt: keep up with new technology, or get left behind.
The result is a first: a fully AI-generated track that reached the Hot 100 before anyone with authority over the chart had ruled on whether it should be there. It was not the earliest sign, either — the AI country project Breaking Rust topped Billboard’s Country Digital Song Sales chart in 2025, on a smaller, sales-only chart.
The takeaway: AI music did not sneak past the rules — it arrived before the rules existed.
Are AI-generated songs allowed on the Billboard charts?
Right now, yes. As of August 2026, no major chart — including the Billboard Hot 100 — has a published rule barring AI-generated tracks. Charts count streams, sales and airplay, and those systems do not ask how a recording was made. A framework has been proposed by record labels, but no chart has formally adopted it.
That gap is what the industry is scrambling to close. On 29 July 2026, a coalition of eleven music companies — Sony Music, Universal Music Group, Warner Music Group, Believe, BMG, Concord, Dirty Hit, Glassnote Records, HYBE, Mom+Pop Music and Partisan Records — published a set of principles they want charts worldwide to apply.
The six conditions the labels proposed
- Any generative AI service used must be properly authorised and lawful
- The recording must be substantially human-made
- It must not raise stream or chart manipulation concerns
- It must comply with copyright, related rights and personality rights
- Releasing it must not breach the AI service’s own terms
- AI use must be disclosed to listeners on streaming platforms
Read carefully, these rules would not ban AI from the charts. A mostly human record that used a licensed AI tool for one element, and said so, would still qualify. A fully synthetic track built on an unlicensed model would not. The sticking point is the phrase “properly authorised” — the coalition’s own members do not fully agree on which AI platforms clear that bar.
How do streaming platforms detect AI music?
Detection does not rely on listening for bad singing. Modern systems analyse the audio file for inaudible spectral signatures that generative models leave behind — statistical fingerprints in the frequency data that survive compression but are inaudible to humans.
Deezer built the best-documented version of this. The French streaming service reports its detector runs at 99.8% accuracy, missing roughly two AI tracks in a thousand and falsely flagging fewer than one genuine track in ten thousand. It has been running since January 2025, and since June 2026 Deezer has shown listeners an explicit AI tag on affected albums.
That technology has spread beyond Deezer’s own app. Billboard and Rolling Stone have both been reported as users of Deezer’s detection tool for identifying AI tracks on their charts — which is how the industry now knows the scale of what is arriving.
Detection is never proof on its own, though. The same limitation applies to AI text detectors, which is why being accused of using AI is so hard to disprove in schools and workplaces. A confidence score is evidence, not a verdict.
The numbers behind the flood
The volume is the part most people underestimate. Deezer’s figures for June 2026 show roughly 90,000 fully AI-generated tracks uploaded per day, representing more than 50% of all new music arriving on the platform. In April 2026 that share was 44%. In January 2025, when detection began, it was around 10,000 tracks a day.
Then comes the twist. Despite dominating uploads, AI-generated music accounts for only 1% to 3% of total streams on Deezer. And in 2025, up to 85% of the streams those fully AI tracks did receive were identified as fraudulent and demonetised.
That combination tells you what most of this flood actually is. It is not listeners embracing synthetic pop. It is streaming-fraud operations uploading generated audio at industrial scale and pointing bot traffic at it to siphon royalties from a shared payout pool.
Spotify’s countermeasures follow the same logic: it pays royalties only on tracks passing 1,000 streams in a rolling twelve-month period, a threshold designed to demonetise the long tail where bulk AI spam lives, alongside stricter verification and song-credit rules.
The takeaway: the AI music problem is mostly a fraud problem wearing a creativity costume.
Can listeners actually tell the difference?
Overwhelmingly, no. In a survey conducted by Deezer with Ipsos, 9,000 people across eight countries — the US, Canada, Brazil, the UK, France, the Netherlands, Germany and Japan — listened to three tracks and tried to identify which were fully AI-generated. Around 97% could not reliably tell.
Their reaction to learning this was equally clear. More than half said the result made them uncomfortable, and 80% said fully AI-generated music should be clearly labelled for listeners.
Labels change behaviour, too. A peer-reviewed study published in Cognitive Research: Principles and Implications in March 2026, by Sarah H. Wu of Stanford University and Kevin J. Holmes of Reed College, found that listeners engaged less deeply with music presented as AI-made — even when the track was actually human-made. Expectation, not audio quality, drove the response.
This is why disclosure has become the industry’s preferred lever rather than an outright ban. If people cannot hear the difference but care deeply about it, the honest fix is a label, not a filter.
What this means for listeners and musicians
If you mainly listen to music, the practical change is small but real. Expect AI tags to appear on more platforms, expect chart entries to carry disclosure eventually, and expect at least a few songs you already like to turn out to be partly generated.
If you make music, the calculation is sharper:
- Disclosure is becoming the price of eligibility. Under the proposed principles, an undisclosed AI element is what disqualifies a track — not the AI itself.
- The platform you use matters more than the output. “Properly authorised and lawful” targets the model’s training data and licensing, which you cannot hear in the result.
- Human contribution needs to be documented, not just claimed. Credits, stems and session files are becoming the evidence trail.
The same shift is happening across creative work generally, where clients increasingly pay for judgement and direction rather than raw output — which is exactly what the market for freelance AI skills already reflects.
Frequently Asked Questions
Can a fully AI-generated song win a Grammy or chart at No. 1?
Charting is currently possible, as “Rubberz” demonstrated by reaching No. 58 on the Billboard Hot 100. Awards are separate: major award bodies generally require meaningful human authorship for a work to be eligible, so a fully synthetic track faces a much harder path to a trophy than to a chart position.
Do AI-generated songs earn streaming royalties?
They can, but platforms are narrowing the gap. Spotify pays out only on tracks that reach at least 1,000 streams in a rolling twelve-month period, and Deezer demonetises streams it identifies as fraudulent — which covered up to 85% of streams on fully AI tracks in 2025.
How accurate are AI music detectors?
Deezer reports 99.8% accuracy for its own system, with roughly two misses per 1,000 AI tracks and under one false flag per 10,000 genuine tracks. Accuracy varies widely between tools, so a single detector result should be treated as evidence rather than proof.
Is it legal to release music made with AI tools?
In most markets, yes, provided you hold the rights to what you release and do not imitate a real artist’s voice or likeness without permission. Copyright protection for purely machine-generated output is limited in several jurisdictions, including the United States, where human authorship is required for registration.
Will streaming services start labelling AI music automatically?
Deezer already does, having added listener-facing AI tags in June 2026. Other platforms are moving toward disclosure requirements rather than automatic detection labels, and the labels’ July 2026 proposal would make disclosure a condition of chart eligibility.
