AI Music and Royalties: What Changes the Pool?

Separate AI upload counts from eligible listening and royalty income. Ten checks for labels, with dated platform sources and a simple allocation example.

More AI-generated uploads do not, on their own, tell a label how much royalty income it will lose. Uploads, listening, eligible activity and money allocated to rightsholders measure different things. A change in one cannot simply be substituted for a change in another.

To assess an AI-related royalty claim, identify the service, rights category, market, period and calculation being discussed. Then separate the platform's reported findings from what your own source reports establish. The ten controls below keep that distinction visible.

What royalty dilution means in a defined pool

Spotify describes streamshare as a monthly calculation within each country. Its eligibility guidance limits the recording-royalty calculation to qualifying activity. A catalog's share depends on its eligible listening relative to the listening counted in that pool, and on the amount available for allocation.

That is different from dividing money by the number of tracks uploaded. An upload with no qualifying listening does not add listening to this denominator. Changes in the pool's revenue or the catalog's own activity also matter. For the wider model comparison, see pro-rata and user-centric royalties.

A small example: uploads versus listening

Assume a fictional recording-royalty pool of EUR 1,000, 100,000 equally weighted eligible plays, and 1,000 plays for catalog A. There are no other allocation weights or exclusions.

• Initially, A receives 1,000 / 100,000 × EUR 1,000 = EUR 10.
• Adding uploads with no eligible plays changes neither the assumed pool nor the counted plays: A still receives EUR 10.
• In a separate scenario, other recordings attract 10,000 additional eligible plays while A's plays and the pool stay fixed. A receives 1,000 / 110,000 × EUR 1,000 = about EUR 9.09.

These are deliberately fixed assumptions, not a Spotify tariff, a Deezer model or a forecast of AI's impact. Real revenue, eligibility and listening can change together. Nothing in the example identifies how the other recordings were made. The amount allocated to a rightsholder is also separate from the artist's contractual royalty, balance and bank payment.

Ten controls for evaluating AI-related royalty changes

1. Keep upload share and listening share separate

Record the unit and denominator beside every percentage. In its 20 April 2026 announcement, Deezer reported roughly 44% of daily uploads as fully AI-generated, alongside 1–3% of total streams. Its 21 July update reported upload share above 50% at peak in June. These are dated Deezer observations, not today's industry totals or percentages of royalties lost by your catalog.

2. Distinguish music creation from manipulated listening

AI use describes aspects of production; artificial streaming concerns listening that does not reflect genuine user intent. Spotify's artificial-streaming policy excludes confirmed artificial activity from royalties. A production technique or an unusual chart is not that confirmation. Use the label investigation guide when there is a notice or unexplained activity to investigate.

3. Record the service and right affected

Spotify's recording-royalty eligibility rule requires at least 1,000 streams in the preceding 12 months plus an undisclosed minimum of unique listeners. Eligibility is assessed monthly. Its guidance says this calculation does not change publishing royalties. Do not apply that recording-side rule to every service or composition payment. Keep a dated policy register with the applicable source and scope.

4. Check identifiers alongside rights records

Use recording and work identifiers to locate the relevant records. IFPI explains that an ISRC stays with a recording when ownership changes. An ISWC identifies a musical work, not its recording or ownership shares. Check participant names, agreements, territories and effective dates separately. Preserve the original value and the reason for any metadata correction; an identifier is not a clearance certificate.

5. Retain the actual AI disclosure

The original September 2025 Spotify announcement described work on AI disclosures through DDEX. Spotify's 20 October 2026 update says AI Credits launched in April and are visible in mobile Song Credits. Record what the creator disclosed and what the delivery partner submitted. Confirm support in the particular delivery route; do not infer production history from a missing field or from how a track sounds.

6. Refer uncertain rights questions to qualified counsel

Keep production disclosures, permissions and the executed agreement together. Ask counsel whether the actual deal addresses synthetic vocals, likeness permissions, warranties and responsibility for inaccurate disclosures. This is a recommended review task, not an opinion that any particular AI recording is lawful, infringing or owned by a specified party. A platform label does not resolve those questions.

7. Preserve each confirmed accounting change

Keep the original report, correction, notice, currency, affected period and approval. Distinguish a change to source income from the resulting artist royalty and any separately authorized fee. Do not pass a platform-level adjustment straight into an artist balance without checking the agreement and calculation. For the wider records process, see royalty accounting controls.

8. Compare the same scope before attributing a cause

A lower income figure alone does not identify AI as the cause. Compare the same service, territory, usage period, right and reporting basis. Record changes in catalog coverage, release activity, source-file versions and available breakdowns. Leave missing explanations unresolved. The streaming-count reconciliation worksheet covers ordinary count and money differences.

9. Keep catalog and agreement relationships traceable

For a disputed recording, retain the relationship between the delivery record, recording, release, agreement and payee. Identify the evidence for each relationship instead of treating a matching display name as ownership. A system's configured royalty rule should be checked against the approved interpretation of the agreement; it does not supply that interpretation.

10. Assign investigation and approval responsibilities

Decide who preserves a notice, asks the distributor for an explanation, checks the royalty calculation and approves a supported correction. This guide provides no universal fraud score, automatic withholding rule or escalation threshold. An unresolved concern remains a concern until the evidence supports a more specific conclusion.

A practical review for the next reporting cycle

Use a 30-day review as an optional planning window, not a promise of completion or a contractual deadline. Assign an owner to each task:

1. List relevant platform policies and distinguish current rules from announced changes.
2. Check priority recordings against identifiers, rights records and disclosed AI use.
3. Trace one representative source amount through the agreement to the statement.
4. Confirm how notices, corrections and unresolved questions are recorded.
5. Ask counsel about any unresolved disclosure or rights language.

For each cycle, preserve source files, record the period and rights type, compare like-for-like figures, and keep the evidence behind decisions. Completion means the chosen checks are documented; it does not certify that the catalog is free of infringement or manipulated listening.

What Qlero's statement records can establish

Qlero's calculation guide explains the sales-row royalty base, contract participation and applicable rate. The Sales CSV documentation describes the corresponding source, date, currency and deal fields. The record company can choose and rename columns, so inspect the actual export.

Statement downloads require Artist Portal access to the relevant payee. Total Royalties and the closing balance differ because the statement also accounts for other entries. Neither figure alone verifies a bank settlement. These records support tracing reported amounts; this review establishes no Qlero AI-classification or fraud-detection capability.

Corrections must respect the period lifecycle. Advanced publishes before closing; Simplified publishes when closing. There is no Unpublish action, and a Closed period cannot be reopened or recalculated. A later Other adjustment can correct a balance without rewriting the closed statement. Advance and Other entries use their full amounts without royalty-rate scaling and do not move money. Determine the correct balance difference first.

Frequently asked questions

Does AI music automatically reduce other artists' royalties?

Upload counts alone cannot establish that. The pool, qualifying listening, allocation rules and the catalog's own share are needed. The worked example holds these variables fixed to explain one mechanism, not to predict the market.

How does streamshare affect the comparison?

It makes the relevant listening denominator important. A percentage of new uploads is not a percentage of eligible listening. Do not substitute either for the artist's contractual rate.

Are AI-generated tracks the same as artificial streams?

No. Keep production disclosures separate from evidence about listening intent and confirmed platform action. Investigate a particular case without accusing a creator on the basis of sound or style.

What did Deezer report about AI streams?

Its 29 January 2026 release described fully AI-generated tracks as 1–3% of its streams and said up to 85% of streams on those tracks were fraudulent in 2025, depending on the month. That 85% is not a percentage of all Deezer streams or a measured loss for a particular label. These are the service's findings, not independent tests conducted for this article.

Does Spotify's threshold apply to publishing royalties?

Its stated eligibility calculation concerns recording royalties. Use the relevant publishing report and rules for composition income, rather than transferring the recording threshold across rights.

Can a royalty statement identify AI use?

Do not assume it contains production-provenance data. A reported amount or adjustment is not, by itself, an AI classification. Request the relevant disclosure or provider notice when that is the question.

Which metadata should we check?

Check recording and work references, contributor identities, the actual AI disclosure and the agreements governing the reported income. Missing data should remain an open question rather than becoming an invented clearance or ownership conclusion.

Does exclusion from recommendations mean removal or non-payment?

Treat recommendation, availability and monetization as separate decisions. Deezer's July announcement described future removal of AI tracks used for streaming fraud or unstreamed for at least six months. Its 13 October announcement described a search-exclusion toggle planned before the end of 2026. That announcement is not proof the toggle is already available to every user. Inspect the actual policy and notice affecting the recording.

How should an unusual streaming spike be investigated?

Use the linked investigation guide and preserve analytics, campaign records and notices. First establish the dates and scope. A spike does not establish how the music was made or who caused the activity.

What belongs in an AI-related royalty review record?

Keep the source files, relevant disclosures and agreements, service and rights category, notice, confirmed amounts, unresolved questions, responsible owner and decision date. Keep a source correction separate from its calculated royalty impact and from settlement evidence.

Trace the claim to the report

Start with the exact question: upload growth, eligible listening, platform allocation, artist royalty or cash received. Keep the evidence for that layer visible. Revisit the review when the relevant policy, agreement or source report changes.

Book a Qlero demo to examine a representative report and the contract-to-statement calculation. The features and integrations pages provide further product context; they are not evidence for a particular AI-related finding.

*Primary sources checked on 10 November 2026. Platform statements are attributed to their issuers; the calculation example is synthetic. No platform detection system or AI-related customer outcome was independently tested.*