Hey, Mark here.

There's a thing every good copywriter does without being told:

As you write a line, you picture the person reading it. You watch their face. You see the small nod, the flicker of "that's me," the eyebrow that goes up when you've overpromised. Then you write the next line to answer whatever you just saw.

You're managing a reaction the whole way down the page, one line at a time. The sentences are just how it shows up on screen.

Here's what I've noticed. When those same copywriters sit down to prompt Claude, that instinct disappears. They describe the copy they want and say nothing about the reaction they want.

So the model gives them what they asked for: clean, fluent, reasonable prose that produces no particular response in anyone.

The fix is to put the instinct back in the prompt.

Let me show you what that looks like.

Sugarman called it the slippery slide. The only job of any line is to get the next line read. Every greased step down the page is a small reaction managed: curiosity held, an objection answered before it forms, a "yes" collected so the next "yes" comes easier.

You don't think about it consciously anymore. You've internalized it. You feel when a line is coasting, and you cut it before you could explain why.

That feeling is the actual skill. The words are downstream of it. A copywriter who can name exactly what a reader should feel at the end of paragraph three will out-write one who's just trying to make paragraph three sound good, every time.

So here's the question worth asking. If reaction management is the skill, why do we leave it out of the one place it would help most?

Look at how a normal copy prompt is built. You tell Claude what you're selling. You give it the audience, the offer, the tone, the structure, the points to hit, maybe a swipe sample for voice.

All of that describes the artifact. None of it describes the response.

The model has no choice but to optimize for the thing you actually specified, which is well-formed copy. It writes to the brief. The brief asked for clear, persuasive, on-voice prose, and that's exactly what comes back: clear, persuasive, on-voice, and completely inert.

The reaction was in your head the whole time. You just never wrote it down, so the model never got it.

This is why "make it more persuasive" goes nowhere. Persuasive to what end, producing what feeling, in whom? You know. The model is guessing. Give it the target and the guessing stops.

The move is to spell out the job of each section as a reaction, not as a topic.

A topic-brief says: "Open with the problem. Then explain the mechanism. Then add proof. Then close with urgency." A reaction-brief says what the reader should be thinking or feeling by the end of each of those.

The difference shows up fast.

The structure is identical. The instruction is completely different. The right-hand column gives the model a thing to aim at, and aiming changes every word choice underneath it.

You'll notice the reaction column is harder to fill in. That's the point. Filling it in is the copywriting. You're doing the thinking the model can't do, then handing it a target sharp enough to write toward.

Once the reactions are written, you do two things with them. You brief with them, and you audit against them.

Here's the prompt shape I use. Adapt the reactions to your project; the structure is the part that matters.

The second pass is where this earns its keep. You're making the model mark its own coasting. When it can't point to a line that produces the reaction, that's your signal: that section is describing the thing instead of selling it, and it's where your edit goes first.

It won't catch everything. The self-audit is a metal detector, not an X-ray. But it points you at the soft spots in thirty seconds instead of three reads, and those soft spots are almost always where your edit was headed anyway.

One more thing it does. Reading the model's account of which line carries which reaction tells you fast whether you briefed it clearly. If Claude's read of the reaction doesn't match what you intended, the gap is usually in your spec, not its draft.

This technique has a hard floor, and you should know where it is.

It only works if you can name the reaction. "I want them to feel something" is not a target. "By the end of the lead, a freelancer who's been undercharging for two years should feel a flash of recognition and a little discomfort" is.

If you can't write the reaction down, the method exposes that you don't have a copy plan yet. You have a topic and a hope. That's useful to learn early, but don't expect Claude to rescue you from it. The model can write toward an emotional arc you define. It can't reliably invent the arc for you, because inventing it requires knowing this reader, this offer, and this moment in a way the model doesn't.

Which is the good news, if you think about it for a second. The part of the job that stays yours is the part that was always worth the most.

You were managing the reader's reaction long before Claude showed up. You just did it silently, in your head, one line at a time.

The only change now is that you have to say it out loud. Write the reaction down, hand it over, and make the draft prove it earned each one.

More clicks, cash, and clients,
Mark Masters