Hey, Mark here.
A younger copywriter told me last week that nobody reads long copy anymore.
He said it the way people say the earth is round, like it was settled.
So I asked him to pull up the last long sales page he'd bailed on. He found one, some supplement page, and I asked him where exactly he quit.
He scrolled, found the spot, and read the line out loud. A paragraph three screens down that restated the headline in slightly different words and moved nothing forward.
"There," I said. "That's the line where you quit. It did no work, so you left. The length had nothing to do with it."
The Reader Quits at the First Dead Line
The "nobody reads long copy" crowd forgets that a reader never sees the whole length at once. They see one line, then decide whether to read the next. Length is only how many of those decisions you asked them to make in your favor.
Sugarman said it cleanest decades ago. The one job of any line is to get the next line read.
Every line is a small yes, and the yeses compound or they break.
A page that earns every yes can run four thousand words and feel fast. A page that whiffs on line nine loses the reader at line nine, and it will not matter that lines ten through three hundred were brilliant. Nobody got there.
So "long copy doesn't work" was always the wrong diagnosis. Long copy that earns its length works fine, and it always has.
What people are reacting to is the flood of long copy that earns nothing, and there is far more of that around than there used to be.
I think I know why.
AI Made the Flood Worse
Ask Claude for a long sales page and it will give you one. That's the problem.
The model is doing exactly what you asked. You asked for two thousand words, so it reached two thousand words, and filler is the cheapest way to get there.
A good share of them come back as transitions that connect nothing, restatements that add nothing, and lines that clear their throat before a point already made.
None of that filler earns the next line. Every padded sentence is a place the reader can leave, and plenty of them do. The AI era filled the channel with long copy that earns nothing, and people looked at the length and blamed that.
Cutting Is the Skill Now
For a hundred years the hard part of copywriting was generating, filling the page, finding enough to say and saying it well.
Claude changed which half is hard. Generating is free now.
Anyone can have two thousand words in a minute.
The scarce skill, the one that separates copy that holds from copy that leaks, is deletion. Knowing which of those two thousand words to kill.
That is good news if you can write, because taste is the exact thing the model does not have.
It will generate all day. It will not feel the line that is coasting. You will.
That small irritation at a sentence taking up space without paying for it is worth more now than it has ever been.
The Earn-the-Next-Line Pass
I run one pass on any draft, mine or the model's, before it goes out.
You go line by line and ask one question of each: does this line make me want to read the next one?
If yes, it stays. If it only sits there being true, it goes.
You can hand the first sweep to Claude.
Go through this copy one line at a time. For each line, tell me the specific job it does: does it open a loop, add a new fact, raise the stakes, answer an objection, or move the reader toward the next line some other way? If a line's only job is to restate something already said or to bridge between two other lines, mark it "cut candidate" and show me the copy reading straight from the line before it to the line after, so I can see what is lost. Do not rewrite. Just mark what earns its place and what does not.
It will not be right about everything. It will flag a line you know is load-bearing, and it will miss one that is dead but grammatical.
The point is not to obey it. The point is that it does the tedious first pass in seconds and hands you a marked-up page, and reading its cut candidates one after another is the fastest way I know to feel where a draft sags.
The reader was always going to run this exact pass on your copy, one line at a time. Their version of "cut" is "close the tab."
You want to find the dead lines before they do.
Where This Goes Wrong
Two things are worth flagging before you take a red pen to everything.
The target is an earned page, not a short one. A high-priced offer to a cold, skeptical reader might need four thousand words to close, and every one of them can be pulling weight.
Cut on instinct toward "shorter" and you will carve out the proof, the objection handling, and the slow build a hard sale actually needs.
The test is always the same. Does this line earn the next one? Length never enters into it.
And the model can flag, but it cannot finally judge. It will call a line a cut candidate when that line is quietly carrying the emotional turn of the whole page, because it reads jobs and not feel. Use its marks as a list of suspects, and keep the verdict for yourself.
The Master’s Memo
Long copy never died. The reader quit at the first line that stopped paying its way, the way they always have, and now there is more dead copy than ever for them to quit on.
Generating is the cheap half of the job. Cutting is the half that is yours.
Run the earn-the-next-line pass before you send, because the reader runs it either way.
More clicks, cash, and clients,
Mark Masters
P.S. Control OS goes from $99 to $199 tonight at 11:59pm Eastern, and stays there. If you have been circling it, these are the last few hours.
