The Polished Persona Problem

Here's a question that almost nobody in business is asking: what happens to trust when anyone can produce perfect communication?

We're entering a world where a deepfake of a CEO can deliver a flawless earnings call. Where an AI can write a blog post in anyone's voice. Where a synthetic video of a founder can give a keynote that is indistinguishable from the real thing. The technology isn't theoretical. It exists today. And it's getting better fast.

The instinct of most organizations is to fight this with detection technology. Build better algorithms. Train people to spot fakes. Watermark authentic content. These are reasonable responses. But I think they miss the more interesting problem. Because the real threat of synthetic media isn't that people will believe fake things. It's that people will stop believing real things.

When any piece of content could be fake, the default posture shifts from trust to suspicion. And once suspicion becomes the default, even authentic communication loses its power. The CEO who gives a genuine, heartfelt message to employees has to contend with the fact that an AI could produce exactly the same message. The founder who writes a thoughtful essay on their company blog has to contend with the fact that ChatGPT could write something similar in thirty seconds. The signal-to-noise ratio doesn't just drop because there are more fakes. It drops because the real thing starts looking like a fake too.

This is the problem I want to think about. Not how to detect deepfakes. How to be unfakeable.

And I think the answer is the opposite of what most people assume. Most leaders, most companies, most brands have spent years polishing their communication. Removing rough edges. Editing out pauses and mistakes. Making everything smooth, professional, and on-message. This used to be the right strategy. In a world where production quality signaled competence, polished communication was a competitive advantage.

But polished communication is exactly what AI is best at producing. Ask any language model to write a professional email, a corporate blog post, or a press release, and it will give you something clean, coherent, and perfectly structured. It will sound like every other professional communication you've ever read. And that's the problem. When AI can produce perfect polish on demand, polish stops being a signal of quality. It becomes noise. It's the baseline. Everyone has it. And when everyone has it, it communicates nothing.

What AI cannot easily produce is the specific, particular, idiosyncratic way that a real human communicates when they're not performing. The half-finished thought that reveals how someone actually thinks. The slightly awkward phrasing that proves someone wrote it themselves. The tangent that goes somewhere unexpected. The opinion that's too specific to be algorithmically generated.

I think this is why certain kinds of communication have started to feel more trustworthy than they used to. Warren Buffett's annual letters have always been famously unpolished. He writes the way he talks. He uses folksy metaphors. He makes self-deprecating jokes. He occasionally contradicts himself between paragraphs. None of this is accidental, and none of it could be easily replicated by an AI, because the idiosyncrasies are the signature. The roughness is the proof of authenticity.

Compare this to the average Fortune 500 CEO's public communications. Press releases written by a PR team. Blog posts drafted by a ghostwriter and reviewed by legal. Social media managed by an agency. Earnings call scripts rehearsed for hours. Every word is optimized. Every edge is sanded. The result is communication that could have been produced by anyone, which means it could have been produced by anything, including an AI. And increasingly, the audience knows this.

I think we're approaching a moment where imperfection becomes the most valuable signal in communication. Not fake imperfection, not a calculated "relatability" strategy where you deliberately include a typo or an "um" in your podcast. People can detect that too. Real imperfection. The kind that comes from someone thinking in public, making mistakes, changing their mind, and letting you see the process.

This applies to brands as much as it does to individuals. The companies that will be hardest to fake are the ones with the most distinctive voices. And distinctive voices, almost by definition, are imperfect ones. They have quirks. They have opinions that not everyone agrees with. They have a specific way of seeing the world that can't be replicated by prompting a model with "write in the style of [brand]."

Patagonia is hard to fake because their communication consistently does things that a rational marketing team would never recommend. Telling customers not to buy your product. Suing the government. Giving away the entire company to a climate trust. These aren't polish. They're the opposite of polish. They're the kind of decisions that only make sense if you actually believe what you say you believe. And that's why they work as trust signals. They're costly to produce and impossible to simulate.

Basecamp's communication is hard to fake for a different reason. Jason Fried and DHH have spent years writing in ways that are deliberately opinionated, occasionally abrasive, and frequently at odds with Silicon Valley consensus. You couldn't train a model on their writing and produce something that feels like them, because what makes them distinctive isn't their prose style. It's the willingness to say things that a model optimizing for broad appeal would never say.

I think the pattern here generalizes. In a world where any surface can be synthesized, depth becomes the only reliable signal. And depth, by its nature, is messy. It contains contradictions. It has rough edges. It doesn't look like a press release.

The practical implication for founders and leaders is counterintuitive. Your best defense against a world of synthetic media isn't to make your communication more polished. It's to make it more human. Write your own stuff. Let the thinking show. Have opinions that are specific enough to be wrong. Use language that an algorithm wouldn't choose. Let people see you work through an idea in real time instead of presenting the finished version.

This won't prevent someone from making a deepfake of you. But it will make the deepfake feel wrong to anyone who knows how you actually communicate. Your imperfections are your fingerprint. They're the one thing that a model trained on generic communication cannot replicate, because they're the things that make you specific rather than general.

The old advice was: be professional. Clean up your message. Remove the rough edges. Present a polished front.

The new advice, I think, is closer to the opposite. Be so specifically yourself that you become impossible to imitate. Because in a world where perfect is free, perfect is worthless. And the only thing that can't be faked is the thing that was never polished in the first place.