28 APR 2026 · 7 MIN READ
Stop asking AI to 'make it shorter'.Three rewrite prompts that actually work.
The single most overused prompt in 2026 is 'make it shorter.' It is also one of the worst. Three replacements I use weekly.
Of all the AI prompts in common use in 2026, "make it shorter" is the most popular and the worst. Founders use it on emails. Writers use it on first drafts. Product teams use it on changelogs. The output is universally worse than the input — shorter, yes, but flat, hedge-stripped, and weirdly impersonal.
"Make it shorter" optimises for the wrong thing. The goal is rarely shortness. The goal is the thing the user actually means but won't say: make it land harder. Make it sound like a real person. Cut the fluff but keep the spine.
Three rewrite prompts I use every week, in place of "make it shorter." Each does a specific job. None of them ask for length.
Prompt 1: The busy-founder rewrite
This is the one that does 80% of what people mean when they ask for "shorter."
Rewrite this for a busy founder skimming on their phone between two meetings. Keep every fact and number. Cut every adjective. Replace passive voice with active. End with the single most important sentence on its own line.
Input: [paste]
Three things make this work. First, the audience is named — "busy founder skimming on their phone" is a specific reading mode, not a generic "shorter". Second, the constraints are operational — keep facts, cut adjectives, active voice, isolate the punchline. Third, the final sentence on its own line forces the AI to commit to a single takeaway, which is usually what was missing in the first draft.
Use this on: cold emails, customer emails, internal updates, board notes, weekly summaries, anything more than 100 words that has to be read in under 30 seconds.
Prompt 2: The "remove the slop" rewrite
The second most common problem with first drafts is not length — it is AI-flavoured slop. The hedges, the "delving", the unnecessary qualifiers, the corporate-AI tone that has crept into every junior writer's prose since 2023. This is the prompt for that.
Rewrite this in the voice of someone who has done the thing themselves and is slightly annoyed at having to explain it again. Use sentence fragments where natural. Cut every "moreover", "in addition", "it's worth noting". Keep the technical accuracy. Keep specific examples. Sound like a person, not a content engine.
Input: [paste]
The "slightly annoyed at having to explain it again" framing is the load-bearing part. It produces prose that respects the reader's intelligence — no preamble, no over-explaining, no "as we've established" filler. The prose comes out leaner without you ever asking for fewer words.
Use this on: blog posts, landing page copy, anything you wrote first in a hurry and then read back and felt the cringe. The cringe is the slop. This prompt is the antidote.
Prompt 3: The verb-led rewrite
This one is for anything that needs to drive action — a CTA, a job spec, a pricing page, a launch announcement, a product description. Most first drafts of action-oriented copy are noun-heavy and weak. This prompt fixes that in one paste.
Rewrite this so every sentence starts with a verb. Cut weasel words ("might", "could", "potentially", "perhaps") unless they are factually necessary. Replace "we believe", "we think", "we feel" with the underlying claim. Keep all numbers exact.
Input: [paste]
Verbs at the front of sentences make prose feel decisive. "We believe this approach is best" becomes "Use this approach." "There may be opportunities to optimise" becomes "Optimise these three things." It is a small structural change with an outsized effect on tone.
Use this on: any copy with a buy/click/signup at the end of it. Any internal doc proposing an action. Any announcement.
Why "make it shorter" fails
"Make it shorter" gives the AI no information about what to cut. So the AI cuts indiscriminately — usually the specific, surprising, particular bits, because those are the bits that take more words to support. What's left is the generic framing, because the generic framing is short by default. You end up with a draft that is both shorter and more boring, which is the opposite of what you wanted.
The three prompts above each tell the AI what to optimise for: a specific reader, a specific tone, a specific structural pattern. Length falls out as a byproduct, but it is never the goal. The goal is sharper, not shorter.
Bonus prompt 1: the "what's missing" inversion
Sometimes a draft is too short, not too long. The default fix is "expand this", which yields the same problem in reverse — flabby, generic expansion that adds words without adding ideas. This is the inverse prompt:
This draft is missing something. Without rewriting it, list the three specific points that would make this 30% more useful to a [audience]. For each, say what kind of evidence would make it credible — a number, a story, a counterexample, or a quote.
You then go and find the evidence. The AI doesn't make it up. The draft gets meaningfully better, not just longer.
Bonus prompt 2: the headline test
Most rewriting fails because the writer never decided what the piece was actually about. The result is a draft that hedges between two ideas. This prompt forces the decision before any rewriting happens.
Read this draft. Give me three competing one-line headlines. Each headline must be a specific claim, not a topic. Then tell me which headline the body of the piece is actually arguing for, and which sentences in the body undermine that headline.
The AI will usually surface a structural flaw — the headline you wanted to write versus the one you actually wrote. Cut the undermining sentences. The piece sharpens immediately. No rewrite needed.
Bonus prompt 3: the objection inversion
This one is for landing pages, sales emails, and anything that asks for a decision. Most copy makes the case for buying. It rarely engages with the case against. The result reads as marketing rather than a conversation.
Read this copy. Now play the smartest possible sceptic — the version of the reader who is one click away from closing the tab. List the three strongest reasons they would not buy. Then rewrite the three weakest sentences in the copy to address those objections directly, without becoming defensive.
The "without becoming defensive" clause is critical. Defensive copy ("we know what you're thinking, but…") is its own slop. The good version of objection-handling reads as confidence, not apology.
Why "make it shorter" became the default
Worth a brief pause on this. "Make it shorter" became the default rewrite prompt for a reason — it works often enough that it feels like a tool. The early versions of GPT-3.5 cut surprisingly cleanly. People learned a habit, and the habit stuck through three model generations even as the underlying behaviour changed.
What modern frontier models do well is follow specific instructions. What they do badly is guess what you actually want from a vague directive. "Make it shorter" sits squarely in the second category. The three prompts above sit in the first.
This is the meta-pattern across most "my prompts stopped working" complaints: the user is treating the model like an early-2023 GPT that needed a magic-word incantation, rather than a precise tool that does what it's told. Tell it what you want, specifically, and it will do that thing. Ask it to "make it better" and it will move three commas around.
The chained version (for high-stakes copy)
For anything you're going to put real money or real reputation behind — a launch announcement, a fundraising email, a public landing page — chain three of the rewrites in a row rather than running each in isolation.
The order I use is this. First, the headline test (bonus prompt 2) to find what the piece is actually arguing for. Second, the busy-founder rewrite (prompt 1) to compress without losing the spine. Third, the objection inversion (bonus prompt 3) to stress-test the result against the smartest sceptic.
Run sequentially in a single chat. Paste the result of step one into step two, the result of step two into step three. The output of the chain is consistently better than any single prompt run alone, because each prompt is correcting a different category of weakness in the previous draft. It takes about ten minutes for a 600-word piece. Worth it for anything that has to land on the first read.
One thing not to do
Do not paste the rewrite prompts as raw instructions to a junior writer or a colleague. They are calibrated for a model — specifically, for the way frontier models respond to constrained directives. A human reader will find them clinical and mildly insulting. The rewrite prompts above are for the AI. The human conversation about why a draft is weak is a different artefact entirely, and a more delicate one.
Where the rest of these live
The rewrite prompts above are part of the cross-cutting prompt library that runs through every Atlas bible. The fullest applied collection — copy rewrites for product pages, lead-magnet drafts, sponsor outreach — is in Bible 02 (Digital Publishing Empire) at £47 launch price.
If you want a free preview, the free chapter from Bible 02 includes the busy-founder rewrite in the email-marketing section.
And if you only ever take one thing from this post: stop typing "make it shorter." Type "rewrite this for [specific reader] [specific tone] [specific constraint]." The output will be unrecognisably better. Same number of words to type.