Taylor Swift’s Latest Move Against AI Fakes Is Clever, But Probably Won’t Work

Taylor Swift has been dealing with AI copycats for years now. Deepfake songs, fake endorsements, voice clones peddling scams — you name it, someone’s tried it. Her latest move? Filing trademark applications for two phrases she says in a promo for her new album: “Hey, it’s Taylor Swift” and “Hey, it’s Taylor.”

On the surface, this sounds like a smart way to weaponize IP law against generative AI. If the USPTO grants the trademark, Swift’s team could theoretically go after anyone using those exact phrases in a way that confuses people into thinking it’s really her. That includes AI-generated audio that mimics her voice saying those words.

But here’s the thing: trademark law is about protecting consumers from confusion, not protecting celebrities from impersonation. The legal standard is whether someone hearing “Hey, it’s Taylor Swift” would believe the speaker is actually the singer. If the context makes it obvious it’s a fan parody, a commentary, or even an AI experiment, that confusion might not exist. And without confusion, there’s no infringement.

Swift’s team is essentially trying to trademark a common conversational opener. That’s a heavy lift. The USPTO generally frowns on granting exclusive rights to everyday phrases unless they’ve acquired a very specific secondary meaning in commerce. “Hey, it’s Taylor” is pretty generic. Even “Hey, it’s Taylor Swift” could be argued as a standard greeting from a public figure.

That said, the filing includes audio clips from an actual promotion, which strengthens the argument that these phrases are tied to her commercial identity. It’s not just her saying them in a random interview — it’s a branded, promotional use. That gives the application a bit more weight.

Still, I’m skeptical this will stop the worst AI abuses. The real problem isn’t someone making a deepfake that starts with “Hey, it’s Taylor Swift.” It’s the fake songs, the manipulated videos, the voice clones used to trick fans into sending money. Those attacks don’t need those specific phrases. They can start with “Hi, this is Taylor” or just launch straight into a song.

What Swift is doing here is more of a signal than a solution. It tells the industry she’s paying attention and willing to use every tool available. But the legal system is playing catch-up, and trademark law wasn’t designed for voice cloning. It was designed to stop someone from selling knockoff handbags.

I’ve seen this pattern before with other celebrities — they try to trademark their name, their catchphrase, their likeness. It works for merchandise and endorsements. It doesn’t work for AI-generated content that exists in a legal gray zone. The courts are still figuring out whether a synthetic voice can even infringe on a trademark when the underlying technology is the one doing the “speaking.”

Swift’s team knows this is a long shot. But they also know that every legal filing, every public move, builds a paper trail. If and when a major AI impersonation case goes to court, having a trademark on record could make the difference between winning and losing.

For now, this is a smart PR move dressed up as a legal strategy. It buys time, puts bad actors on notice, and keeps the conversation going. But the real fix will have to come from legislation or court rulings that directly address AI impersonation — not from trademarking a greeting.