LANDR’s 95% figure

LANDR says Spotify flagged 95% fewer artificial streams for music it distributes in the second quarter of 2026 than in the first quarter of 2025. The distribution and music-creation platform is also expanding fraud detection developed with machine-learning lab Mila, according to Music Business Worldwide’s September 24 report.

That is a substantial claimed reduction, and a welcome direction for anyone putting a record into circulation without manufacturing its audience. The useful work starts with the unit being counted. LANDR is talking about artificial streams flagged by one service across its distributed catalog. An artist choosing a delivery partner will want to translate that aggregate result into something closer to release-day experience.

Think of the meter on a noisy return channel. A lower reading invites you to check what changed upstream. The equivalent checks here involve total listening volume and the definition of a flag.

Put the count beside the rate

A flagged-stream total describes an amount of activity identified by the platform. A flagged-stream rate would compare that amount with total Spotify streams for LANDR-distributed music over the same period. Both are useful, and they answer different questions.

For a hypothetical distributor, flagged streams falling by half while total streams also fall by half would leave the flagged share unchanged. That example says nothing about LANDR’s actual volumes. It shows why an artist assessing a distribution service needs the listening total beside the detection total.

The 95% claim describes the size of a decline between two reporting periods. Establishing the proportion of streams flagged in either quarter would require those separate totals.

A useful reporting view would show counts and rates on the same basis, with the intervening quarters included. That would reveal whether the decline happened gradually or arrived in one abrupt step.

Two systems, one statistic

The measurement chain has two distinct participants. LANDR is describing its detection work. Spotify supplies the flags being counted. A drop in Spotify’s flagged total is compatible with improvements at LANDR, but the percentage alone cannot establish how much of the change LANDR’s system caused.

Catalog composition matters. A different set of releases changes what is being measured. Comparing a stable group of recordings alongside the whole catalog could help separate screening effects from a shifting roster. Consistent flag definitions matter too. Any change in the yardstick during the period would need to be explained before the figures could be compared confidently.

Mila’s involvement identifies the research collaborator. The reduction figure alone tells us nothing about the system’s inputs or error rates. The label AI fraud detection also leaves open what the system examines. Audio, account behavior and delivery activity would each raise different questions about the decision being made.

Give false alarms a review path

Aggregate improvement can coexist with errors on individual releases. For a musician whose activity is questioned, the pressing concern is how to obtain an explanation and request a review. A release campaign has dates attached to it. An unresolved warning can leave the person running that campaign unsure what to do next.

A useful notice would identify the affected recording and time window, describe the concern in ordinary language, and explain what context the musician can submit. Publishing every detection threshold could make a system easier to evade. Basic communication can still give an artist a specific issue to address.

A review might consider campaign dates or links to coverage. Those materials offer context without proving that every play was genuine. The musician should also be able to check the review’s status without repeatedly sending the same explanation.

These are criteria for evaluating a distributor. LANDR’s percentage alone tells us nothing about how many legitimate releases were challenged or how quickly disputed decisions were resolved.

Keep the campaign receipts

The practical move is small enough to fit into export day. Alongside the final master and delivery notes, keep a folder for the release campaign. Name it with the release title and date.

  • Save the starting point. Capture the analytics available before promotion begins and retain the dates. A baseline helps explain what changed without relying on a memory of how the dashboard looked last Tuesday.
  • Log the promotion. Record ad dates, creator posts, playlist outreach and coverage. Save correspondence with anyone hired to promote the release, especially their explanation of how listeners will be reached.
  • Check spikes against the log. Compare unusual movement with those dates, using whatever source or geographic detail the service supplies. A spike is a reason to look for context before deciding what happened.

If someone offers a guaranteed play count while staying vague about the audience, ask how the campaign works before paying. Keep the answer. If a distributor later contacts you, preserve the notice and your replies in the same folder.

This should stay lightweight. Artists cannot audit every listener or control everyone who shares a track. The aim is a usable record of your own actions, with enough context to answer a specific query.

Before the next delivery

LANDR’s expansion belongs on a working musician’s radar because distribution is part of the release workflow. Earlier, more accurate screening could make that workflow less disruptive. The company’s reported decline gives artists a specific result to examine, with scale and error handling kept in view.

A helpful follow-up to the percentage would show comparable quarterly rates and explain the handling of disputed decisions. These can be presented without publishing a detector’s playbook. Musicians would gain a way to compare the service on something closer to everyday reliability.

Before your next delivery, ask your distributor what information a review would require and how status updates are provided. Save the answer beside the release plan. If a warning arrives during launch week, you will have a starting point that is more useful than refreshing the dashboard.