Blue Planet III is leaving the plankton in.

You know what I’m talking about if you’ve ever watched nature documentaries. Those little floaty bits that drift across the frame. The slightly out-of-focus moment. The shot that’s good but not perfect. For decades, BBC’s Blue Planet series defined itself by crystal-clear footage that looked almost too good to be real. Now they’re doing the opposite on purpose.

Executive Producer Elizabeth White told The Guardian they’re keeping “the amount of plankton and things” that previous productions would have cleaned up. They’re going “warts and all.” The goal? To prove it’s authentic. To prove it’s real.

Think about that for a second. We’ve reached a point where perfection reads as fake.

This isn’t just a quirky production choice. It’s a tell. Blue Planet III is responding to a world where AI-generated content has gotten so good that polish itself has become suspicious. When things look too seamless, too flawless, we assume a computer made them. So now, if you want people to trust that humans actually filmed your nature documentary, you need to show them the mess.

It’s the same anxiety driving Lee Chang Dong’s approach to filmmaking. His new film Possible Love engages with class and labor without being didactic, keeping the human messiness intact rather than smoothing it into a clean narrative. Different medium, same instinct: authenticity requires imperfection.

And here’s the thing. They’re right.

The authenticity tax

This is what I’ll call the authenticity tax: the extra work creators now have to do to prove they didn’t use AI. It’s not enough to make something good anymore. You have to make something that looks demonstrably human.

Pennsylvania just passed a law requiring labels on AI-generated ads. Any ad that’s been “significantly modified or generated by artificial intelligence” needs disclosure, whether it’s video, audio, images, or text. That applies to Super Bowl commercials and your local restaurant’s flyer equally.

The law assumes we can tell the difference. That there’s a clear line between “significantly modified by AI” and just, you know, edited on a computer like everything else. Good luck with that.

But the bigger issue is what this does to creative work. Now every choice gets interrogated. Did you clean up that audio too much? Did you color-correct until it looked too good? Did you edit out all the boring parts and accidentally make it feel algorithmic?

The Blue Planet III approach is smart because it gets ahead of this. But it’s also kind of tragic. We’re adding flaws on purpose now. We’re manufacturing authenticity by keeping the mistakes in.

What we actually want

Here’s what gets me: we spent decades building tools to remove imperfections. Noise reduction. Image stabilization. Auto-enhance everything. We wanted the tech to make our work look better, cleaner, more professional.

Now the tech got too good at it, and we’re running in the opposite direction.

This isn’t just about AI skepticism. It’s about what happens when the gap between human capability and machine capability gets narrow enough that we can’t tell them apart anymore. The rough edges aren’t just proof of humanity. They’re proof of effort. Proof that someone was actually there, in the water, fighting with the equipment, dealing with the plankton.

When everything can be generated, the only thing that matters is what can’t be generated. And right now, what can’t be generated is the imperfect. The messy. The real.

The actually interesting question

But here’s where it gets complicated: how long does this last?

Right now, AI-generated content tends toward the smooth and the perfect because that’s what the training data rewarded. Clean images rank higher. Polished videos get more views. The models learned to optimize for that.

But what happens when we start training models on intentionally imperfect content? When “authentic-looking plankton drift” becomes a feature instead of a bug? When the AI learns to generate the rough edges too?

Then we’re back where we started, except now we don’t even have imperfection as a reliable signal.

Blue Planet III’s strategy works today. It might not work in five years. The authenticity markers we’re using right now, the tells that something was made by humans, those aren’t permanent. They’re just the current edge of the capability gap.

And that gap is closing.

What this means for everyone else

If you’re making anything right now, any kind of creative work, you’re dealing with this whether you realize it or not. Your audience is already trying to figure out what’s real. They’re looking for the seams. They’re hunting for evidence of AI.

You can either ignore that and let people wonder, or you can do what Blue Planet III is doing: make your humanity obvious. Keep the rough edges. Show the work.

But understand what you’re trading. You’re trading polish for trust. You’re accepting a ceiling on how good your work can look because looking too good makes people suspicious.

That’s the world AI has created. Not one where machines do all the work, but one where humans have to work harder to prove they did any work at all.

The plankton stays in the shot. Not because it looks better, but because it looks real. And right now, that’s worth more.