Edit your own writing so it sounds like you meant it

Reading your own draft immediately after finishing it is close to useless, and the reason is mechanical rather than a matter of willpower. You still have the intended meaning loaded in your head, so your eyes pass over the sentences and your mind supplies what you meant. A sentence that is genuinely ambiguous reads as clear, because you know which of the two readings was intended. A missing step in an argument reads as present, because you made the connection while writing and it has not yet faded.

The cheapest fix is time. An hour helps, a night helps considerably more, because what you are waiting for is the intended meaning to stop being immediately available so that you read the words instead of your memory of them. Where a deadline makes that impossible, the substitutes are ones that break the pattern rather than add effort: read it on a different device, print it, change the font, read it from the last paragraph to the first. None of these are as good as sleeping on it, and all of them are considerably better than a careful second read of the same words on the same screen ten minutes later.

Reading aloud is the highest-yield editing technique available and the one most people skip because it feels theatrical. It works because your voice has stricter requirements than your eye. A sentence too long to say in one breath is too long to read comfortably. A clause that has to be re-read to be understood makes you stumble audibly. A change of register — a paragraph of plain speech followed by a sentence of borrowed formality — sounds obviously wrong when spoken and passes silently when scanned.

The instruction is not to read carefully; it is to notice where you falter and mark it without stopping to fix anything. Every stumble is a genuine problem in the text, even when the cause is not immediately obvious, and marking is faster than diagnosing. Then go back through the marks. This works especially well for anything that will be read on a phone or spoken by a person — a message, a script, a short piece of copy — because those are consumed at close to speaking pace. For anything conversational in tone, it is the difference between writing that sounds like a person and writing that sounds like a template.

First paragraphs are usually the writer clearing their throat. They set up context the reader does not need yet, restate the title in longer form, or explain why the topic matters before saying anything about it. This happens for a legitimate reason: writing the first paragraph is how many people work out what they think, so it records the process of arriving at the piece rather than the piece itself. That makes it valuable to write and usually unnecessary to keep.

The test is to delete it and read from the second paragraph. Surprisingly often the piece starts better — and it starts better because the second paragraph is where the writer stopped preparing and began saying something. If deleting it leaves a genuine gap, the missing information is normally one sentence rather than a paragraph, which can be folded into what follows. The same applies to the beginning of individual sections and to the opening line of an email. "I hope this message finds you well. I wanted to reach out regarding…" is two sentences of throat-clearing before the subject arrives, and a reader on a phone may not scroll past them.

Three mechanical passes catch a large share of what makes prose sludgy. First, filler: just, really, very, quite, actually, basically, simply, in order to, at this point in time. Search for them and delete unless the sentence loses meaning, which it rarely does. These accumulate because they are how people speak, and speech has tone of voice to carry emphasis while text does not — so in writing they weaken rather than intensify. "This is really very important" is less forceful than "this matters".

Second, passive voice, which is not wrong but is often evasive. "The delay was caused by a supplier issue" hides who is responsible; "our supplier delivered late" does not. Convert where there is an actor worth naming, keep it where the actor is genuinely unknown or irrelevant. Third, long sentences, and the mechanical version of this pass is to look for "and" joining two complete thoughts and replace it with a full stop. Do not do it everywhere — uniformly short sentences read as staccato and slightly aggressive. The aim is variation, with the longest sentences no longer than one comfortable breath.

For every sentence, ask what the reader would lose if it were deleted. Then actually delete the ones where the honest answer is nothing. This is a harder instruction to follow than it looks, because the sentences that fail this test are frequently the ones that took longest to write and read most smoothly. A well-turned sentence that carries no information is more likely to survive editing than a clumsy one that carries some, and that inversion is why most drafts stay too long.

Applied at the paragraph level it is even more useful and more uncomfortable. Whole paragraphs of business writing exist to signal thoroughness rather than to inform: the paragraph explaining why the topic is important, the one summarising what was just said, the one anticipating an objection nobody has. Cutting them makes a piece shorter and denser, and dense is what a reader deciding whether to keep going responds to. The related habit is to distrust your favourite sentence. It is not always bloat, but it is the sentence most likely to have been written for your own satisfaction rather than the reader's benefit, and it deserves the test more than the others.

Self-editing has a point of diminishing returns and then a point of active harm. Past a certain number of passes you begin making changes that are lateral — swapping one adequate word for another, reordering clauses that were fine — and each pass adds risk of introducing an error while producing no improvement. The signal that you have arrived there is when a pass produces changes you cannot justify beyond preference, or when you find yourself reversing a change you made in the previous pass.

A workable stopping rule is to run each mechanical pass once — aloud, filler, passive, long sentences, the deletion question — and then stop regardless of how you feel about the result. What you cannot get from any number of passes is the thing you actually need next, which is whether the piece makes sense to somebody without your knowledge. That requires a reader, and a single question to one colleague who was not involved — what do you think this is asking you to do? — will find more real problems than three more rounds of your own attention. Self-editing makes writing clear. It cannot tell you whether it is clear to anyone else.