How do we get AI to sound like our brand?

The scene I'm in this week

This week a founder forwarded me a welcome email his team had just shipped and asked, half-proud and half-suspicious, whether it was any good. I read the first two lines and knew. Not suspected. Knew. A machine wrote this, and almost nobody on his team had touched it after.

I wasn't reading his mind. I was reading the fingerprints. There's a particular sound AI copy makes when it goes out unsupervised, and once you've heard it a few hundred times you can name the song from the first bar. The email opened with "In today's fast-paced business landscape." It had an em dash in nearly every line. Halfway down sat the sentence "It's not just a tool, it's a partner in your growth." It closed with "The future of work is here. Let that sink in."

Four tells in one short email. Each one on its own is nothing. A coincidence. Stacked together they're a confession. And the founder asked me the question I hear constantly: "Can you help us clean this up so it sounds less like AI?"

Yes. But the way he asked it is the whole problem. He wants to edit what came out of the machine. This email, then the next one, then the one after that, forever. He's about to put a person on payroll whose entire job is sanding down robot copy by hand. And he's missing the one move that would make almost all of it unnecessary. Let me name it.

Naming what's actually broken

Here's what's actually broken, and it isn't the copy. The founder believes this is an editing job. Call it the Cleanup Trap: the assumption that you fix generic AI by scrubbing each draft after the fact, one piece at a time, forever.

Here's why the trap is so easy to fall into. The cleanup works. You delete "delve," thin out the em dashes, and the draft genuinely improves. Then you do it again tomorrow. And next week. You've quietly turned a system problem into a chore, and the chore has no end, because you're forever editing the output of a machine you never actually briefed.

Think about what you handed that machine. A blank prompt and a vague ask. It knew nothing specific about your company, your buyer, your fight, or your point of view, so it did the only thing it could. It reached for the statistical center of everything it was ever trained on and wrote from there. I've been calling that empty space the Context Vacuum for two years. The generic copy, the six tells, the whole anonymous hum of it, that's just the sound of a vacuum getting filled by the average of the internet.

You can't edit your way out of a vacuum. You can only fill it. And the thing that fills it has a name: a Magnetic Messaging Framework. Think of it as a documented brand bible written for the machine, the context layer the AI reads before it writes a word. It carries the two things a model can never invent on its own: your strategic truth, and your writing rules. Hand it those up front, and the cleanup mostly stops being necessary, because the generic version never makes it onto the page.

Why this is worse now than ever

Editing AI output by hand was a tax you could maybe afford when one person made one draft a week. That math is gone.

AI dropped the cost of producing copy to zero. Your team now spins up homepages, decks, emails, and posts at a volume no human editor can keep up with. If your quality control is a person reading every draft and pulling out the delves by hand, you have two moves and both of them lose. Edit everything, and you've made a human the bottleneck on a machine built for infinite scale. Edit nothing, and the internet's average ships under your logo every single day. There's no amount of manual cleanup that keeps pace with the firehose.

Here's the part that should reframe the whole thing. When AI can produce infinite clean, confident, average copy for free, the scarce asset isn't editing skill. It's a specific, lived point of view, written down clearly enough that a machine can carry it. That's the entire argument behind Strategic Positioning Is the Only Moat AI Can't Copy. The framework is how that point of view actually gets inside the machine. Without it, you're just a faster way to produce the same copy as everyone else in your category.

What actually goes inside the framework

A Magnetic Messaging Framework isn't a style guide and it isn't a single clever prompt. It's the brand bible the AI reads before it writes anything. Three things go inside it, and the order they go in matters.

First, your truth. The specific buyer in the specific situation you were built for. The villain you're against. The contrarian belief you'll defend at a conference. The outcome only you deliver. The exact words your last happy customer used on a renewal call. This is the layer that makes the copy sound like YOU, because it's the only thing in the document a model could never have read somewhere else.

Second, your voice. How you actually sound on a good day. Short lines mixed with long ones. Rhetorical questions. Contractions. The analogies you reach for. The register you use with a CEO versus the one you use with your champion. This is the layer that turns "technically our information" into "unmistakably our voice."

Third, the detox layer. This is the part founders are stunned to learn they can simply hand the machine as a standing rule. You don't hunt these tells down after the fact. You forbid them up front, one time, and the model obeys them on every draft after. Here's exactly what goes in it:

1. 1Ban the "it's not X, it's Y" reflex. "It's not a tool, it's a partner." "This isn't software, it's a movement." It's the move AI makes when it wants to sound deep without saying anything. The rule you write down: state the thing in plain, positive terms, no antithesis.
2. 2Cap the em dashes. Real writers use them constantly, so the dash itself isn't the crime. The volume is. One in nearly every other line is the fingerprint. The rule: use them sparingly, and reach for a period or an ellipsis instead.
3. 3Kill the vocabulary nobody says out loud. Delve, tapestry, testament, robust, seamless, leverage, elevate. The rule: every word has to be one you'd actually say to a person across a table. "Delve into" becomes "dig into." "Leverage" becomes "use."
4. 4Break the rule of three. "Unlock, streamline, and elevate your workflow." The model stacks three even when there's one real idea in the sentence. The rule: pick the single verb that carries the weight and cut the other two.
5. 5Delete the windup. "In today's fast-paced world." "In the ever-evolving landscape of..." The rule: open on the actual point. The first sentence should be the one you were going to get to in the third.
6. 6Forbid the motivational landing. "The future is here. Let that sink in." The rule: end on a fact, a specific, or a real stake, never a reach for a feeling.

Read those six again. That's not a checklist you run on every draft until you retire. That's a paragraph you write one time and paste into the framework. The distance between those two things is the distance between a chore and a system. Does that land?

What I see across 100+ B2B companies

The pattern is almost funny. A team learns the tells, gets good at the scrub, and proudly sends me a cleaned-up draft with the delves gone and the dashes thinned out. And it reads like a competent paragraph that still could have come from any of their competitors. They got better at editing, and the copy still sounds like nobody, because editing was never going to put a point of view into a document that didn't start with one.

And the editing never compounds. When researchers started tracking word frequency in scientific papers after late 2022, they watched a quiet word like "delve" go from one almost nobody used to one of the most reliable signs that a machine had a hand in the writing. Teams everywhere now delete that single word by hand, draft after draft, with no end in sight. They're bailing water. Nobody's patching the hole.

What I see across more than a hundred B2B companies is a clean split. The ones still stuck treat this as an editing problem and put a human on permanent cleanup duty behind the AI. The ones who broke out treated it as a context problem, built the machine a framework once, and stopped editing. That's the same root cause behind Why does AI keep producing generic content for our company? and the real cost in What You Lose When AI Content Sounds Like Everyone. The model was handed an empty brief and asked to sound like a company it had never met. The answer was never a better editor. It was a brief worth writing.

A real example

A composite, drawn straight from a pattern I see constantly. A cybersecurity company, around $20M ARR, sharp founder, genuinely differentiated product. They'd been doing the detox the hard way. A style guide that banned "leverage" and "seamless." A rule about em dashes. A junior marketer whose entire job was running every AI draft through a human pass before it shipped. Their content was clean. It was also, in the founder's own words, "fine, and totally forgettable." And it was costing them a full salary to stay exactly that forgettable.

We didn't hand them a better editor or a longer banned-words list. We built them a Magnetic Messaging Framework and turned it into an AI Brand Twin, a custom model that writes with the framework already loaded. We sat with the founder and two of his best engineers and pulled the raw material out of them. The specific way they describe the threat their product kills. The phrase a customer used on a renewal call that made the founder go quiet for a second. The category belief they'll argue at a conference. The villain they're genuinely angry at. We wrote all of it down, the voice rules and the detox layer included, into one document the machine now reads before it writes.

The marketer stopped editing, because there was nothing left to clean up. The framework caught the tells before they reached the page, and the truth was already sitting in the draft. The same AI tool, pointed at the same task, started producing copy the founder described as "weirdly, that actually sounds like me on a good day." The detox didn't get done better. It got done once, and from then on it got done by the system. That's the whole idea behind The AI Brand Twin: Scaling Your Voice Without Losing Your Soul.

What this means for you

If your AI copy sounds like everyone else's, you don't have an editing problem and you don't have a prompt problem. You have a missing framework. The model is doing exactly what it's built to do. With no brief about you, it hands you the average of everyone, tells and all.

Stop sanding the output. Knowing the six tells is genuinely useful, but running them by hand on every draft is a treadmill you never get off. The leverage is building the machine a Magnetic Messaging Framework one time: your truth, your voice, and the detox rules, in a single document the AI writes from on every job. One of those is a chore you repeat forever. The other is a system you build once. Which one do you want to own?

Here's what to do this week:

1. 1Name the real problem out loud. Every hour your team spends manually de-robotting AI drafts is an hour spent bailing instead of patching the hole. The hole is that the machine has no framework of yours to write from.
2. 2Write down the three layers. Your truth (the villain, the contrarian belief, your buyer's exact words), your voice (how you actually sound), and your detox layer (those six rules, written once). Together, that document is your Magnetic Messaging Framework. It already exists in your head and in your customers' mouths. It's just not on paper yet where a machine can read it.
3. 3Put it in the machine, not in your editor's memory. Load the framework into your AI tools as a standing brief, or build a Brand Twin that writes with it loaded, so every draft starts from your truth and self-corrects the tells. That's the moment the cleanup stops being anyone's job.

The detox everyone obsesses over is real, and it's worth doing exactly once: as a few lines inside a framework, not as a chore you repeat on every draft until you retire. Give the machine your truth and your rules up front, and it stops handing you the internet's average to fix. It starts handing you back yourself. This is just truth.