Stop Asking AI To Write Your Why. Ask It To Attack It.
By Atiba de Souza
The Videos That Gave It Away
Every candidate we screen gets the same interview questions, answered on video, before anyone talks to them live. It is how we hear the person before we meet them.
The videos started coming back identical. Not similar. Identical. Same structure, same phrasing, same rhythm of an answer that had clearly never lived in that person's mouth before. One candidate left the AI's own closing line pasted at the bottom of a written answer: the model politely offering to draft follow-up responses.
Nobody was answering our questions. Everybody was answering the same prompt.
That is not a hiring story. That is every business owner right now, feeding their "why we do this" to AI and publishing what comes back. If you would spot it instantly in a job candidate, you have to spot it in your own content. The tell is the same: the words are fluent, correct, and could have come from anyone.
The Framework Underneath This
The content that actually moves a stranger from knowing you exist to trusting you enough to buy breaks into three types, in a specific order and a specific ratio: WHY, then HOW, then WHAT.
WHY is awareness. One pain point, tied to one of three things people actually want: relationships, status, or money. Why it matters, why it's worth solving. The ask is small: raise your hand, answer a question, self-identify. This is 70% of what you publish. WHY is where you find out whether a stranger recognizes themselves in what you said.
HOW is credibility. You showing the mechanics of how you actually think, not the highlight reel of what you concluded. This is 15%. The decision HOW forces on you is different from WHY's decision. WHY gets attacked for being generic. HOW gets attacked for being vague. The question isn't "could a competitor say this," it's "could a stranger actually follow this, or are they just supposed to be impressed by it." A HOW piece that only makes you look smart has failed at HOW even if it never plagiarizes anyone. Publish it only if someone could take the reasoning and do something with it.
WHAT is the pitch. The offer, 15%. The decision WHAT forces is about who's in the room, not what's on the page. WHAT only works on someone who already raised a hand on your WHY or nodded along to your HOW. Send it cold, to someone meeting you for the first time, and you've skipped the two steps that earned the right to ask. So before you publish anything labeled WHAT, the real question is not "is this pitch good," it's "has this person seen enough of me yet that a pitch doesn't feel like a stranger asking for money."
Seventy percent of what you put in front of strangers should be WHY. Not the offer, not the features, not the price. Most businesses run that backwards and then wonder why nobody self-identifies with them before the pitch shows up.
WHY carries the most weight in this system, and it's also the content people most want AI to write for them, because it's the hardest to write and the most exposed. That is exactly the wrong place to hand the pen away.
The One Word This Turns On
Stress-test: attacking your own already-written WHY with AI to find where it's generic, borrowed, or indistinguishable from what the next guy in your industry would say. Not asking AI to write the WHY for you.
Those are two different jobs. One produces content. The other produces a candidate video with the AI's signature still on it.
The Fishing Room You Actually Belong In
I teach AI for a living and I turn down implementation work on purpose. Most people selling AI education are quietly using the teaching to sell you the doing, which is why free teaching so often reads as a trick: you sense the upsell before it arrives. Refusing that revenue is what makes what I teach believable.
Different rooms teach different ways of fishing. The skill you actually need decides which room you belong in, and right now the room most business owners walk into is "have AI write my content," when the room they need is "learn to write my own point of view, then put AI in the ring against it." Those are not the same skill, and they are not taught in the same room.
If someone is offering to hand you back a finished WHY, they are handing you a fish. Stress-testing is the rod.
What Angie's Fix Actually Proves, And Doesn't
Angie Etheridge owns a pharmacy, seventeen years running, and she came to a workshop opening a wellness clinic. She had already done real work: she'd written her business's brain document, the thing that captures her actual point of view, her voice, her differentiation. But every time she used AI on the business, it came back sounding like nobody. She assumed her prompting was broken.
It wasn't the prompting. The projects were never pointed at the brain file. She had written the real thing and then built on top of empty folders. The moment the folder got attached and the instructions named the file explicitly, she produced a vetting guide, a coaching skill, and a launch checklist inside one hour, and all of it sounded like her.
Here's what that story proves, and what it doesn't. It proves the wiring problem is real: AI can only sound like you if you point it at something that actually is you. It does not prove that writing that something is easy. Angie's brain document was seventeen years of running a pharmacy compressed into a page, and getting it onto that page was its own separate, uncomfortable act, done long before any workshop. Naming your actual point of view, tying it to what people really want, and writing it in words that are unmistakably yours is not a checklist you clear in an afternoon. It's the hard, exposed work that most people reach for AI to avoid doing at all.
This is the ranking, plainly stated: writing your own WHY from your own lived position is the whole game. Stress-testing it is valuable but secondary, and it only works on something real. Get the order backwards, either stress-test a hollow document or let AI author the document from scratch, and you've built the candidate-video problem straight into your own marketing.
Name What This Costs You
Nobody wants to hand their WHY to an adversarial AI and hear "a competitor could publish this word for word." That's not feedback on a paragraph. It reads as feedback on your judgment, your years in the business, the thing you believed made you different. It's identity-level, and identity-level critique is the resistance you actually have to push through, not a workflow you skip because it's uncomfortable.
The cheaper, faster path is always there: ask AI to write the WHY from scratch, publish what's fluent, move on. It feels like progress because something got produced. It is the pharmacy-clinic problem before the fix, and the interview-video problem after it. Fast and wrong, dressed as done.
| Inconsistent application | Consistent application |
|---|---|
| Front-load WHAT this week because the offer feels urgent | Hold the 70/15/15 ratio even when you're impatient to sell |
| Stress-test the WHY when you remember, skip it under deadline | Run every WHY piece through the attack before it goes out, no exceptions for being busy |
| Publish HOW content that impresses but can't be followed | Publish HOW content specific enough that someone could actually replicate the thinking |
| Send WHAT to a cold stranger who's never seen your WHY | Send WHAT only to someone who already raised a hand on your WHY or HOW |
| Blame AI for generic output some weeks, work around it other weeks | Check every single time whether AI can actually see your real point of view before you judge what it hands back |
The plain instruction row matters more than it looks. Saying exactly what you want out of a relationship, out loud, costs you nothing and removes the other party's guesswork. AI defaults to agreeable the same way a new hire defaults to telling you what you want to hear. If you don't say "attack this, don't validate it," you'll get validation, and validation is not a stress test.
Where This Doesn't Apply
If you haven't written a real WHY yet, don't stress-test one. There's nothing there to attack. Do the Angie work first, and expect it to take the kind of time and discomfort hers took: sitting with what actually makes you different, saying it plainly instead of hedging it into something safe, and getting it into a document AI can actually see. That's not a mechanical step you clear before the "real" work starts. It is the real work. Stress-testing a document that was never written from a real position just polishes emptiness until it's fluent emptiness, and fluent emptiness is exactly what showed up in those interview videos.
The Monday Move
Take the WHY piece you're about to publish. Paste it into AI with one instruction, stated as plainly as you'd state it to a person: "Attack this. Tell me exactly which competitor of mine could publish this word for word and nobody would notice the difference." Not "give feedback." Not "improve this." Attack it.
Expect it to fail the first time. Expect it to fail the second time too. Two or three rounds of rewriting is normal, sometimes more, especially the first time you run this on a WHY you've never actually examined before. Each round, you're not polishing a sentence, you're sitting again with "a competitor could say this," and that doesn't get easier to hear just because you already heard it once this week. That discomfort is the process working, not the process malfunctioning.
If it survives, publish it. If it doesn't, you just found out before your audience did, which is the entire point of a stress test. The alternative is finding out the way we found out with the interview videos: after it's already gone out, wearing someone else's signature.