Skip self-editing and send a manuscript to beta readers after one pass, fix the obvious typos, cut the scene that already felt slow, call it done, and you can end up with feedback like this: two readers stuck in the same chapter, for reasons that turn out to be the same reason, a third flagging a character motivation that makes complete sense to the writer and to nobody else, because it lived entirely in their head and never made it to the page.

None of that is a typo. All of it is fixable, and all of it needs fixing before anyone else reads the book.

That’s the job of self-editing: catching the problems a reader can’t unsee once they’ve hit them. Polishing sentences until they shine comes later. Do it before beta readers, and their feedback tells you things you couldn’t have caught yourself. Skip it, and their feedback tells you what a first read-through would have told you anyway, at the cost of their goodwill and your credibility.

Self-Editing Is Three Separate Jobs

The mistake I made at first was approaching self-editing without a structure. I’d read through the manuscript, editing as I went: fix obvious mistakes, polish a sentence, add a detail that felt missing. Then, partway through, I’d realise large sections needed a full rewrite, and all the tweaks and polish I’d just spent hours on were wasted effort. I’d carefully polished work I was about to cut.

This made it even harder for me to kill my darlings, to make the necessary changes. And it took me a surprisingly long time to realise there was a better way.

Split it into three passes, done strictly in order, and that stops happening.

Structural pass. Story and character. Does every scene earn its place? Does the pacing hold? Is the point of view consistent? This is the only pass where you’re allowed to cut or rewrite whole scenes, because everything downstream depends on the structure being settled first.

Line pass. Sentence-level craft. Rhythm, word choice, dialogue, the difference between showing something and stating it. This is where filter words, passive voice, and weak words get cut.

Technical pass. Grammar, spelling, consistency, formatting. Last, always last, because polishing a sentence you’re about to cut in the structural pass is wasted effort you’ll do twice.

The Structural Pass

Read the whole manuscript in as few sittings as you can manage, ideally without stopping to fix anything. You’re not editing yet. You’re diagnosing.

Does your protagonist’s want and need hold up all the way through, not just at the start? If you can’t say in a sentence what they’re chasing and what they need to learn, structural problems downstream are almost guaranteed, because everything else is meant to be in service of that.

Does every scene earn its place? It should move the plot, change a relationship, or reveal character. If a scene isn’t doing at least one of those, it’s not pulling its weight, however nice the writing is. This is the hardest cut to make, and the one most worth making.

Is the pacing even, or does it drag somewhere you already suspect? If you’re honest with yourself, you probably already know roughly where the manuscript sags. The structural pass is where you stop avoiding that chapter and fix it.

Do your plot threads resolve? Every one you opened needs to close, or be deliberately left open, not just forgotten because you moved on to a different subplot three months ago and never came back.

This is also where you catch point of view slips, timeline inconsistencies, and chapter endings that don’t pull the reader forward. All of it is story-level work. None of it is about sentences yet.

The Line Pass

Now, and only now, you’re working at sentence level.

Read it aloud. This catches more than any silent read-through will. Anything that snags your tongue will snag a reader’s attention too, just less consciously.

Cut the filter words. Felt, saw, realised, noticed. These words put a narrator between the reader and the moment, when most of the time you want the reader in the moment directly. Filter words has the full list and the exceptions worth keeping.

Check your passive voice. Passive voice has real uses. The goal here is making it a deliberate choice rather than a habit. Passive voice in fiction covers when it’s doing you a favour and when it’s just softening a sentence that would hit harder active.

Go through your adverbs. Some adverbs earn their place. Check each one against whether it’s propping up a verb that should be doing more work on its own. Adverbs in fiction walks through which ones to keep.

Hunt down your weak words. Very, really, just, quite, somewhat. Small words that dilute a sentence without adding anything a reader will miss if they’re gone. Weak words in fiction is the full list I check my own drafts against.

Listen to your dialogue. Cover the character names and see if you can still tell who’s speaking from the words alone. If you can’t, the voices aren’t distinct enough yet.

This pass is slower than the structural one and it should be. You’re touching almost every sentence in the manuscript.

The Technical Pass

Last, because there’s no point proofreading a paragraph you might cut. This pass is grammar, spelling, consistency (names, timelines, invented terms), and formatting.

Run a grammar tool if you use one, then read it manually anyway, tools miss things and flag things that aren’t wrong. Build or update a style sheet as you go: character names and their spellings, place names, any invented terminology, key dates. Future you, three books from now, will be grateful you kept one.

One thing worth knowing here: readability tools like Flesch Reading Ease are useful during this pass as a way of spotting sentences that have quietly become unreadable while you were deep in the story. They’re a diagnostic, not a target, fiction isn’t trying to read like a safety manual.

Finish with a proofread in a different format from the one you drafted in. Printed, or on an e-reader. Your eyes skim familiar formatting; changing it forces a proper read.

Writing Desk tip: Before a big structural change, branch the chapter with Chapter Iterations rather than editing over the original. You can compare the new version against what you had side by side, and roll back cleanly if the cut doesn’t work.

Get a Second Read Before Your Beta Readers Do

Self-editing catches what you can see. The harder problem is catching what you can’t, because you wrote every word of it and you’re too close to tell anymore whether a scene lands or just feels like it should.

The Writing Desk AI Editor is built for that gap. It works chapter by chapter, and gives you structured feedback across four lenses:

Developmental. Structure, pacing, and whether the chapter is doing its job within the wider story, the same questions your structural pass is asking, from a second angle.

Beta reader. How the chapter lands for a reader coming to it fresh: what’s engaging, what’s confusing, where attention might drift.

Line edit. Prose-level feedback, the same territory as your own line pass, filter words, passive constructions, adverbs, rhythm, but applied consistently across every chapter rather than whichever ones you had energy left for.

World editor. Consistency, cross-referenced against your own resource sheets, so a character’s eye colour, a timeline, or an invented rule doesn’t quietly drift between chapter four and chapter nineteen.

For each chapter, it highlights specific strengths and weaknesses rather than a generic score, so you know what to look at instead of just knowing something’s off. If a note doesn’t make sense, or you want to push back on it, there’s a chat function to interrogate the feedback further, ask why it flagged something, or argue your case.

What it won’t do is write anything for you. It never generates or touches your prose, every suggestion is yours to take or leave, and the manuscript stays entirely yours, sentence by sentence. Think of it as a second pair of eyes for the moments you don’t have a beta reader to hand: fast, structured feedback whenever you need it. It gets your manuscript to the point where a human reader’s feedback is about the story, not about problems a self-edit should have caught first.

Knowing When to Stop

Self-editing doesn’t end when the manuscript is perfect. It ends when you’ve stopped finding problems you can see clearly yourself. That’s a different milestone, worth being honest with yourself about.

If you’re still finding structural issues, keep going, beta readers shouldn’t be doing your structural pass for you. If you’re mostly finding sentence-level nitpicks and you keep changing the same paragraph back and forth, you’re probably done with what you can catch alone. That’s different from the manuscript being finished. It means you’ve reached the edge of what self-editing can do, and what’s left needs eyes that aren’t yours.

That’s the point where a developmental editor earns their fee, because there’s a real ceiling on what any writer can see in their own work, however many passes they do. It isn’t a sign your self-edit failed. Self-editing gets the manuscript ready for other eyes. It was never meant to replace them.

Final Words

Three passes, in order: structure, then line, then technical. Skip the order and you’ll redo work. Skip a pass entirely and your beta readers will find what you should have found first, at a real cost to how useful their feedback can be.

I’ve got the whole thing as a checklist you can work through page by page, print, or bookmark and tick off as you go.

Get the Self-Editing Checklist →

FAQs

How many self-editing passes do I need?

Three, minimum: structural, line, technical, done in that order. Some writers split the line pass into two (a rhythm pass and a word-choice pass), which is fine if it helps, but structure always comes first and technical always comes last.

Should I self-edit before or after beta readers?

Before. Beta readers are there to catch what you can’t see yourself, plot holes that only show up to a fresh reader, a character motivation that’s clear in your head but not on the page. If you send them a manuscript full of problems you could have caught alone, you’re spending their attention on things a self-edit would have handled.

How long should self-editing take?

Longer than you think, shorter than you fear. For a full-length novel, budget several weeks if you’re doing all three passes properly, not a weekend. Rush it, and you’ll likely be back to your beta readers for a second round you didn’t need.

Do I need a developmental editor if I’ve self-edited properly?

Depends on your goals. A thorough self-edit, run through the AI Editor’s developmental and beta reader lenses, plus honest feedback from actual beta readers, catches a lot. A developmental editor is a further step up from that: a professional who catches structural and craft issues that are difficult to see from inside your own manuscript, however many passes or tools you’ve run it through. Worth reading what a developmental editor does before deciding either way.

Related reading: