You finished the draft. Then you opened it again, and somewhere between the notes, the half-fixed chapters, and the third rewrite of page one, you landed on what one writer in a Facebook group called "Revision Planet": a place you can't seem to leave.
Here's how to revise a novel without living there. Work in six passes, biggest changes first, and give every pass one question and one exit test. You only move on when the pass meets its test, and you stop revising when a full read turns up nothing bigger than a one-scene fix. That's the whole method. The rest of this guide shows you each pass, the tests, and a complete test manuscript going through all of them.
Revising changes what your story does. Editing changes how your sentences read. Revise first. Revision works on the book, the chapters and the scenes: plot, character arcs, structure, what each scene is for. Editing works on sentences and words: clarity, grammar, consistency, typos.
The order matters for a boring, practical reason. Every hour you spend polishing a paragraph is wasted if that scene gets cut in the next pass. Professional publishing runs in the same order: a developmental editor looks at the story long before a copyeditor looks at commas.
Revision works on the top three levels. Editing starts where revision stops.
So when this guide says "revise," it means the big work. Copyediting and proofreading come after, and they're a different job (often a hired one).
The most common advice in every revision thread is also the most ignored: put the draft away before you touch it. Not because rest is magic, but because familiarity is a real handicap. In a 1993 proofreading study, people caught fewer errors in essays they had written themselves than in unfamiliar ones, and the researchers put that down to familiarity (Daneman and Stainton, Reading and Writing). When you know what a page is supposed to say, you read what you meant, not what's there.
So rest it until the story feels a little strange to you. Then read the whole thing in as few sittings as you can, and don't fix anything. Not a typo, not a name. This read has exactly two outputs: a reverse outline and an issue log.
A reverse outline is an outline you write after the draft, describing what's actually on the page instead of what you planned. The idea comes from academic writing (the Purdue OWL teaches it as noting the topic of each paragraph in the margin), but for fiction the unit that matters is the scene.
For every scene, fill one card:
A card from our test manuscript. The "What changes" line is the one most writers leave blank, and it's the one that matters most later.
The "What changes" line does a lot of quiet work. If you can't fill it, the scene probably doesn't turn, and you'll deal with that in Pass 3. If you plotted before drafting, compare these cards against your original novel outline: the gaps between plan and page are your first list of story problems. If you're a discovery writer, the cards are the outline you never had, and they're the fastest way to see where the draft drifted (our guide to story mapping for discovery writers goes deeper on this).
A to-do list says "fix Tomas." An issue log says what's wrong, where, and how far the fix spreads. One line per problem:
That last column is the one most revision guides leave out, and it's what the next section is about.
Every revision guide says "work from big to small." Few say why, and without the why it's easy to cheat: you tell yourself you'll just fix this one clunky paragraph first.
Here's the why. A note's ripple is the number of scenes you'd have to change to fix it. A name misspelled in Chapter 5 has a ripple of 1: you fix it where it sits. A promise the book makes and never keeps has a ripple of 4 or more: the setup, the reminders, the payoff and the ending all have to move.
High-ripple fixes rewrite pages. Low-ripple fixes polish pages. If you polish first, the big fixes will rewrite the pages you polished, and you'll polish them again. That loop is a big part of why revision feels endless.
In our test manuscript, The Salt Road, a courier carries a sealed letter she's told to open only once she reaches the city. She never opens it. Fixing that touches four scenes: the briefing, the night she's tempted to open it, the delivery, and the ending. Meanwhile, the villain's name changes from Venn to Vane in the last chapter. That's a ripple of 1. If we'd fixed the name first, nothing would be lost. If we'd polished the delivery scene first, the letter fix would have torn it up.
Once your log has ripple counts, sort every note on two questions: how far does the fix spread, and how much would a reader feel the problem?
Sort by ripple and reader impact. The top-right box is your Pass 1.
This is also the answer to "where do I even start?" You start in the top-right box. Nothing else.
Here's the whole plan in one view. Each pass asks one question, produces one output, and ends when one test passes.
The six passes. When a later pass turns up a story-level problem, it goes back to Pass 1, not into the line edit.
Then the manuscript leaves revision and goes to editing: a copyedit and a proofread.
One rule makes the map work: a pass only fixes its own kind of problem. If you're in Pass 3 and discover the midpoint doesn't work, you don't patch it in place. You log it, and if it's high-ripple, you go back to Pass 1 for that part of the book. That's what the loop-back arrows mean.
The question: does the book keep the promise it makes?
Every novel makes promises in its opening chapters: a question will be answered, a person will change, a danger will arrive. Pass 1 checks that each one gets kept, and that the road from promise to payoff makes sense. This is where you cut, move and add whole scenes, and it's where the plot, the character arcs and the stakes get their real test.
Work on the reverse outline first, not the manuscript. It's far easier to move a card than a chapter. Rewrite the outline until it works, then make the manuscript match it.
Check these, in this order:
Trey Parker and Matt Stone, the creators of South Park, guest-speaking at an NYU class filmed for mtvU, gave a rule that works beautifully on a reverse outline: between any two story beats, the word should be "therefore" or "but," never "and then" (Writer Unboxed has a good write-up).
Read your cards aloud with those words between them. "Halvard holds the gate, therefore Mara escapes, but he dies" is a story. "They ride, and then they reach the lake, and then they cross" is a list of events. Every "and then" marks a spot where cause and effect broke.
Discovery writers often finish with a draft that wandered off its own premise. That's normal, not a failure. The reverse outline shows you the book you actually wrote, and sometimes that book is better than the one you planned. Decide which promise you're keeping, then cut or rewrite the scenes that serve the other one.
You're done with Pass 1 when the revised outline reads as a chain of "therefore" and "but" from start to finish, and the one-paragraph pitch you'd give a friend matches the book on the page. Not perfect prose, not tidy scenes. Just a story that holds together on paper.
The question: do the facts still agree after Pass 1?
Many revision guides leave continuity to the copyeditor at the very end. That's backwards. Pass 1 creates continuity errors. Cut a scene where a character learns a secret, and every later scene where they act on it breaks. Move a chapter, and the "two days later" in the next one is now wrong.
So continuity gets its own pass, right after the story settles and before you spend any time on scenes. The tool for it is a fact sheet: one place where the truth of your book lives.
That last category saves you from yourself. If a character lies on purpose, log it, or a tired reread will "fix" it. A Random House copyeditor, interviewed for Penguin Random House's author news site, described keeping a detailed timeline and character notes for exactly this work, especially across a series. For more on keeping cast details straight, see our guide on how to keep track of every character.
We planted eight continuity errors in The Salt Road, a five-chapter fantasy story of 3,463 words that we use as a test manuscript, then ran it through several AI tools. The full method and screenshots are in our test of the best AI for novel continuity checking. Two results matter for revision.
Each column is one planted error. Physical details got caught often; timeline, knowledge and promise errors slipped past most tools.
First, surface details get caught most. Eye color got caught by four of the five tools. "Only a day" in a three-day journey was fully caught by just one. The name change, what Tomas knew about the ferryman, a dead man nodding at a trial and the unopened letter were each caught by two. (Two of the tools ran on free tiers only, and a general-purpose model with a careful prompt caught all eight.)
Second, and this is the part we didn't expect: two of our eight "continuity errors" weren't really continuity errors. A courier who hasn't set foot on a boat since her brother drowned, stepping onto a ferry "without a second glance" isn't a wrong fact. It's a character arc that skipped its most important scene. And a letter that's never opened is a broken promise. In the ripple system, both belong in Pass 1. That's why continuity can't be your only check, and why story comes first.
A series turns your fact sheet into a series bible. Everything above still applies, plus three extra columns:
When the new book contradicts an old one, you have three choices: change the new book (usually best), change the old book (possible for self-published ebooks you can update, impossible for copies already sold), or explain the change inside the story. Pick one deliberately and log it. A duology that grows into a trilogy is the classic case: book two's ending was written as a final ending, so the fact sheet has to separate what's truly settled from what can reopen. For fantasy series with big casts and maps, our guide on how fantasy writers stay organized covers the bible itself.
A fact sheet for one book becomes a series bible with three extra columns.
You're done when the fact sheet has no open contradictions. Every conflict is either fixed or logged as intentional.
The question: does every scene turn?
Now the story holds together and the facts agree, so it's finally safe to work inside scenes. A scene turns when something is different at the end: a decision made, a secret out, a relationship cracked, a plan ruined. Your reverse outline already has a "What changes" line for every scene. Pass 3 is where you go through those lines one by one.
For each scene, check:
A scene that doesn't turn has three options: give it a turn, merge it into a neighbor, or cut it. In The Salt Road, the scene where Mara empties her satchel and looks in the mirror changes nothing. It's an inventory with a view. It's also where we planted three of our continuity errors, and in real drafts that's typical: a scene nobody needs is a scene nobody rereads closely.
You're done when you can name what changes in every scene, in one short sentence, from memory.
The question: what do readers feel, and where?
Bring in beta readers after Pass 3, not before. Earlier, they'll spend their attention on problems you already know about. Later, it's too late to act on anything big. Tell them what you want: where they got bored, where they got confused, what they believed, what they didn't. Ask for reactions, not fixes.
Neil Gaiman put the core rule better than anyone: "When people tell you something's wrong or doesn't work for them, they are almost always right. When they tell you exactly what they think is wrong and how to fix it, they are almost always wrong" (his rules for writers in The Guardian).
That rule turns a messy pile of feedback into a sorting job:
Several readers reacting at the same spot is a symptom. Your job is to find the cause, which is often earlier in the book.
Feedback that changes the story goes back through the map: a high-ripple note returns to Pass 1, and anything Pass 1 changes sends you through Pass 2 again.
You're done when every repeated note is resolved and every single note has a decision (fix, find the cause, or decline with a reason).
The question: does each sentence sound like you?
This is the pass most writers want to start with, and the one that finally earns its turn. Read aloud, or use text-to-speech so you hear what you wrote. Listen for:
Not every repeat is a mistake. The phrase "against her ribs" appears three times in our test manuscript, always about the sealed letter. That's a refrain. Cut it, and the letter loses its weight. Line editing is judgment, not search-and-destroy.
You're done when your edits start reversing your earlier edits. You change "said" to "murmured," and on the next read you change it back. That's not a sign you need another pass. It's the sign the pass is over. The choices left are taste, and taste doesn't converge.
Advice is easy. Here's the method on a complete test manuscript, with real counts. The Salt Road is our five-chapter, 3,463-word test story: a courier, Mara Quell, carries a ledger exposing a corrupt city watch across three days of road, guided by a mapmaker with a secret and chased by the watch captain's men. It's short, so the counts are small, but every note below comes from our issue log.
Pass 0, Cold read. 14 scenes, 14 cards. One scene had an empty "What changes" line right away: the satchel inventory in Chapter 4.
Pass 1, Story: 4 notes.
Pass 2, Continuity: 8 errors, 2 intentional. Six are planted errors; the other two (the watch arriving too fast, the wheelhouse) we hadn't planted, and the general model found them. The other two planted errors moved to Pass 1. The eight: eye color, the villain's name, the compass, "only a day," what Tomas knew about the ferryman, a dead man nodding at a trial, the watch arriving too fast after the pigeon, and a wheelhouse on a tiller barge. Seven had a ripple of 1; the timing problem had a ripple of 2. Two more items went into the log as intentional: Tomas's lie, which the story confesses, and a two-week jump between chapters.
Pass 3, Scenes: 2 notes. Merge the satchel scene into the one before it. Cut Tomas's "Three days, like I said" recap on the causeway, but keep the route detail: it sets up the washed-out shore road.
Pass 4, Readers. We didn't run a beta round on a test story, so this pass is method only.
Pass 5, Line: 3 notes. Vary "looked." Cut three of seven "the way." Keep "against her ribs."
Few notes, big reach: Pass 1 had half as many notes as Pass 2 but reached 8 scenes to Pass 2's 5.
Two things stand out. Pass 1 had half as many notes as Pass 2 and the most reach: 4 notes reaching 8 of the 14 scenes. Do those last and you'd rewrite more than half the book you just polished. And scenes that don't turn get the least attention on rereads, which is a good reason to run Pass 3 with suspicion.
Writers ask how long revision takes, and the honest answer is that it takes as long as the passes take. No reliable data gives a number, and the figures you'll see online are mostly one author's habit stated as a rule. What we can say is this: revision deserves real time. The rest between drafting and Pass 0 is part of the work, Pass 1 is slow on purpose, and rushing any pass just moves its problems into the next one.
Two habits keep it from becoming endless:
Sometimes you don't have that time: a preorder date, a contest deadline, a publisher's schedule. That's when a tool earns its place. It can't make the story decisions for you, but it can take the slow mechanical work off your plate:
AuthorFlows keeps the AI optional: the Story Map, Timeline and Character profiles work as a plain place to hold your reverse outline and the character and timeline parts of your fact sheet, whether or not you ever press an AI button.
You stop when the passes stop finding big problems. Not when it's perfect, because it never will be, and not when you're sick of it, because that happens long before it's done.
A concept, not data: each pass changes less than the one before. The flat end of the curve is where you stop.
Three signs tell you the book is ready to leave revision:
Seth Harwood, writing on Jane Friedman's site, describes signs like circular feedback and micro-editing, and suggests a "future book" file for late ideas that don't belong in this one. Use it. Half the reason writers can't stop is a great new idea, and a file keeps it without reopening the book.
When the signs line up, the book goes to editing: a copyedit for consistency and correctness, then a proofread. If you're comparing AI help for that stage, our AI developmental editor comparison covers what the tools can and can't do.
Writers worry about two things with AI in revision: that it will flatten their voice, and that it gets things wrong with total confidence. Both worries are fair. Here's the honest map:
If you want the full step-by-step on using AI this way, our guide on how to self-edit your novel with AI goes through it pass by pass.
Everything above, as one list you can work down. Each pass has its question first and its exit test last.
The whole method on one page.
Pass 0, Cold read: what did I actually write?
Pass 1, Story: does the book keep its promise?
Pass 2, Continuity: do the facts still agree?
Pass 3, Scenes: does every scene turn?
Pass 4, Readers: what do readers feel, and where?
Pass 5, Line: does each sentence sound like you?
Then: copyedit and proofread.
There's no right number, and the figures you'll see online are mostly individual habits. Count passes instead of drafts. A novel is ready when each revision pass has met its exit test and a full read turns up only small, one-scene notes. For one writer that takes three drafts; for another it takes eight. The passes matter more than the count.
Revise in place if your reverse outline shows the story mostly works and the problems are fixable scene by scene. Rewrite from a blank page if Pass 1 changes the premise, the main character's arc or the point of view, because then most scenes would change anyway. A quick test: if more than half your scene cards need new "What changes" lines, a rewrite is usually faster.
Revise first. Revision fixes what the story does: plot, arcs, structure and scenes. Editing fixes how the sentences read: clarity, grammar and consistency. Editing first wastes effort, because revision cuts and rewrites the pages you polished. Leave copyediting and proofreading until every revision pass has met its exit test.
A reverse outline shows you the book you actually wrote, scene by scene, instead of the book you planned. For each scene you note what happens, why it's there, whose point of view it's in and what changes by the end. It makes plot gaps, scenes that don't turn and drifted storylines visible at a glance, and it's much faster to rework than the manuscript itself.
AI tools can help with parts of revision, such as summarizing scenes, flagging continuity contradictions and spotting repeated words. They can't decide what your story should be, and letting them rewrite your prose tends to flatten your voice. Use AI to find problems and do the mechanical checks, then make the decisions and the rewrites yourself.
Start with the reverse outline, not the sentences. Long drafts usually carry whole scenes that don't turn, repeated beats and subplots that serve a different promise. Cutting those in Pass 1 and Pass 3 removes far more words than trimming lines ever will. Line-level trimming in Pass 5 then takes care of the rest.
Manage characters, plotlines, and relationships visually. Start using AuthorFlows now!
Get Started Now