A Play Count Is Not an Audience

Key takeaways

• Generating music with AI and using bots to inflate streams are separate activities.
• Fake plays can divert royalties from legitimate listening when payouts depend on a shared revenue pool.
• A play count alone cannot distinguish real fans from automated activity.
• Platforms need trustworthy listening records, clear payout rules, and a way to challenge mistakes.

A song can rack up plays without building an audience. The AI music streaming fraud case involving Michael Smith raises an awkward question for streaming platforms: when software supplies both the tracks and the plays, what exactly are they paying for?

The money follows the records

Royalties compensate rights holders for the use of their music. In streaming, deciding which plays count toward payment can shape who gets paid.

Consider a service that divides a fixed revenue pool according to each track’s share of eligible streams. If fabricated plays enter that calculation, they claim a share of the same money available for legitimate listening. Other rights holders can end up with less.

Actual payment systems and contracts vary. There is no universal formula in which every play earns a fixed amount.

But every system that uses listening records faces the same dependency: the records must be trustworthy. Perfect arithmetic cannot rescue a payout calculated from fake activity.

Cheap supply meets manufactured demand

AI music generation handles the supply side: producing tracks. Bots can handle the activity side: repeatedly playing those tracks without someone pressing play each time.

Connect the two, and someone can attempt to collect royalties without first attracting listeners. The catalog grows. The counters move. An audience remains optional.

The crucial distinction is between creating music and manufacturing demand for it. A music file proves that a track exists. A stream counter, by itself, does not prove that anyone wanted to hear it.

That distinction also keeps the debate about AI music focused. Whether a song was generated by software is a different question from whether its apparent audience was fabricated.

Your favorite song on repeat complicates things

Suppose a track records 100 plays.

Those plays might come from one devoted listener, a hundred people sampling the song, or a program running it repeatedly with nobody listening. The total looks the same.

That makes repetition a tricky signal. Plenty of people play one song all afternoon. A system that treats that behavior as proof of manipulation risks penalizing ordinary listening.

Platforms therefore need to look beyond totals, examining account activity and playback patterns together. The challenge is to identify artificial activity while correctly recognizing legitimate repeat listening.

A large number on a dashboard is easy to understand. Establishing what happened behind it takes more work.

An AI label answers only half the question

An AI label tells listeners something about how a track was made. It says nothing about who played it or why.

AI-generated music can attract real fans. Human-made music can receive bot traffic. Music provenance and listening authenticity need separate checks.

That gives us a more useful way to assess a platform’s response. Which plays does it exclude from royalty calculations? How can a creator challenge an incorrect decision? What happens when manipulation is discovered after payment?

Those rules matter beyond any individual fraud case. Enforcement addresses a particular scheme; clear payment and appeals procedures help protect the next royalty distribution.

Streaming turns recorded listening into compensation. As both music production and playback become easier to automate, platforms need to establish whether a real listener sits behind the record. Our subscription money follows those records, so their credibility matters.