# Novel I — Revision: Structural Edit
In ["Novel I — First Draft: Word Count Targets"](/harrys-desk/novel-i-first-draft-word-count-targets) I said the first draft is a search, not a performance. Today the search is over. You have a stack of pages — maybe 80,000 words, maybe 50,000, maybe 120,000 — and the question changes. The question is no longer "what happens next?" The question is "what did I actually write?"
That question is harder than it sounds. You did not write what you outlined. No one does. The outline was a hypothesis. The draft was the experiment. Now you are reading the results, and the results do not match the hypothesis in ways that are sometimes better and sometimes worse and always different. The structural edit is the work of reconciling the book you planned with the book that exists, and choosing — for the first time — the book you are actually going to have.
This is the most important pass in the life of a novel. It is also the pass most writers skip.
Why Writers Skip the Structural Edit
The first draft is exhausting. By the time you type the last sentence, you have lived inside the story for months. You know every character's secrets. You know which scenes were easy and which were extracted under duress. You know the chapter you wrote at two in the morning that you have never reread. The last thing you want to do is read the whole thing cold, from the beginning, with the discipline of a stranger.
So most writers do something easier. They go back to the beginning and start polishing. They fix a sentence here, tighten a paragraph there, improve an opening. This feels like revision. It is actually procrastination dressed in work clothes. Line-level fixes are real and necessary, but they belong to the third pass, not the first. If you polish a scene that should not be in the book, you have wasted the polish. If you sharpen a chapter that is in the wrong place, you have made the wrongness more presentable.
The structural edit comes first because it is the only pass that can ask the big questions: Is this the right structure? Does the book start in the right place? Are there scenes that do nothing? Are there promises the first act makes that the third act never pays? Is the protagonist's arc continuous, or does it reset twice? Are there two characters doing the same job? Is the climax earned?
You cannot answer these questions by fixing sentences. You can only answer them by reading the whole book as if you have never seen it before.
Reading as a Stranger
The hardest part of a structural edit is achieving distance. You wrote the book. You know what every scene is supposed to do. The reader does not have your annotations. The reader has the pages.
I use a simple trick to create distance: I put the draft away for at least two weeks. Not forever. Not for a year. Two weeks is enough for the sentences to lose their freshness in my ear, for the intentions to blur, for the memory of what I *meant* to write to fade slightly behind the memory of what I *did* write. When I come back, I read the whole draft in two or three sittings, as fast as I can, with a pen in my hand and no computer open.
I mark four things as I read. Nothing else. Not word choices, not metaphors, not grammar. Only:
1. Where I was bored. Boredom is diagnostic. It means the scene is not doing enough work, or it is doing work the reader does not need yet, or it is repeating something the reader already knows. 2. Where I was confused. Confusion means I withheld something too long, or introduced a character without enough context, or shifted timelines without signaling. 3. Where I was surprised in the right way. Surprises I did not plan — a character who came alive in chapter nine, a subplot that found its own engine — these are the gifts of the first draft. I mark them so I do not cut them later by mistake. 4. Where I was surprised in the wrong way. A character who acts against their established nature without the reader seeing why. A plot turn that resolves something too easily. A tonal shift that breaks the contract.
This read-through is not a luxury. It is the foundation of every decision that follows. If you skip it, you are editing a memory of the book, not the book itself.
The Structural Map
After the cold read, I build a structural map. This is not an outline — the outline is what you had before you wrote. The structural map is a description of what is actually on the page, scene by scene, in the order the reader will encounter it.
For each scene, I write one or two sentences answering three questions: What changes? Who drives the change? What does the reader know after this scene that they did not know before? If I cannot answer all three, the scene is either filler or a scene I do not understand yet. Both are problems.
The map reveals things the draft hid from you. You see the middle that sags. You see the subplot that started strong in chapter four and then vanished for two hundred pages. You see the character who was supposed to be central and appears in only three scenes. You see the climax that depends on information the reader last encountered in chapter two, four hundred pages ago.
This is the moment where AI becomes genuinely valuable — not as a writer, not as a polisher, but as a diagnostician. I will get to the specific techniques shortly. First, the decisions.
The Four Decisions of a Structural Edit
A structural edit comes down to four categories of decision: cut, move, expand, and restructure. Everything else is a later pass.
Cut
Most first drafts have at least ten percent that does not belong. This is not a failure. It is the nature of discovery. You wrote scenes to find out what the book was about. Some of those scenes did their job by existing, not by staying.
The hardest cuts are the good scenes that do not serve the whole. A beautifully written dialogue, a scene that made you cry, a chapter that captures a relationship perfectly — if the book does not need them, they have to go. I have a document called "The Morgue" where I paste cut scenes. It comforts me to know they are not deleted, merely shelved. But they do not belong in the novel.
How do you know if a scene serves the whole? The structural map tells you. If the answer to "What changes?" is "nothing" or "the reader learns something they already know," the scene is a candidate. If removing the scene does not break any later scene's logic, it is not pulling its weight. Cut it. Do not soften the cut by leaving a summary. If the reader does not need the information, the summary is also not needed.
Move
Some scenes belong in the book but are in the wrong place. A structural edit is when you discover that the flashback in chapter twelve would have more power in chapter six, where the reader is still learning who the character is. Or that the subplot introduction is buried in the middle when it needs to begin earlier to earn its payoff.
Moving scenes is where the novel's architecture becomes visible. You are rearranging load-bearing walls. The structural map — especially the "What does the reader know after this scene?" column — tells you what order the reader needs to receive information in. Sometimes a scene is doing everything right except arriving too late.
AI can help here in a specific, bounded way. Feed the model the structural map and ask: "Which scenes contain information that is used before the reader receives it?" The model will flag chronology violations you may have been too close to see. It will also flag false positives — scenes that intentionally withhold — but the false positives are easy to dismiss. The real catches are worth their weight.
Expand
The structural edit also reveals where the draft is thin. A character who appears in three scenes but carries the emotional weight of the ending needs more presence. A relationship that is supposed to be deep but was never shown being built needs scenes. A turning point that happens too fast needs room.
Expansion in a structural edit is not about adding words. It is about adding scenes. The distinction matters. Adding words to an existing scene is a line-level pass. Adding scenes that change the structure is the structural pass. If the ending depends on the reader caring about a friendship, and the friendship has one scene, you need two more scenes. Not two more paragraphs. Two more scenes.
Restructure
The most ambitious structural decision is to change the architecture itself. This is rare but real. Sometimes the cold read reveals that the book is not the book you thought you were writing. The protagonist is not the person you thought was the protagonist. The A-story is actually the B-story. The chronology that worked in outline feels inert in execution, and the book wants to be told differently.
I do not take restructuring lightly. It is the most dangerous kind of revision because it can feel like starting over, and the temptation to start over is the temptation to never finish. But if the cold read tells you the book's center of gravity is in a different place than you planned, you have to move the structure to match the gravity. Pretending otherwise produces a novel that feels like two books stitched together.
Using AI as a Structural Diagnostician
The model's role in a structural edit is different from its role in drafting. In drafting, the model is a generator of possibilities. In a structural edit, the model is a reader — and specifically, the stranger you are trying to become.
Here is how I use it.
The blind summary test. I paste the entire draft (or chapter by chapter, if the model's context window cannot hold the whole thing) and ask: "Summarize what actually happens in this book, in order, without referencing what the author intended. Describe only what is on the page." The model will produce a summary that is honest in a specific, limited way: it can only report what the text says, not what the text means. That limitation is exactly what I want. If the model's summary of my protagonist's arc does not match my intention, the gap is real. The reader will have the same gap.
The character presence audit. I ask the model to list every named character and count the scenes in which they appear. This takes the model about thirty seconds and takes me half a day by hand. The results are often surprising: the character I thought was the antagonist appears in four scenes. The character I thought was minor appears in twelve. The counts do not lie. They tell me where the draft's actual energy is.
The promise audit. I ask the model: "What questions does the first act raise? Which of those questions does the third act answer?" The model will list the open questions — mysteries, character goals, thematic promises — and then check them against the resolution. Unanswered questions are not always failures. Some are deliberate. But the audit gives me the list, and I can decide which are intentional and which are accidents.
The pacing map. I ask the model to rate each chapter's narrative momentum on a scale of one to five, based on how much the situation changes per thousand words. The model's ratings are not perfect, but the shape of the graph is diagnostic. A first act that averages four and a second act that averages two tells me the middle is sagging. A climax that rates a two tells me the ending is arriving without force.
The key discipline with all of these is the same one I have been preaching since the [Integrity Framework](/harrys-desk/the-integrity-framework-a-writers-code-for-ai-use): the model diagnoses. You decide. The model can tell you that chapter seven has no change. It cannot tell you whether chapter seven is a deliberate breath in the rhythm or a scene that should be cut. That is a judgment about the whole book, and only the writer holds the whole book.
The Pillar Test
After the structural map and the AI diagnostics, I run what I call the pillar test. I identify the four to six scenes that the book cannot survive without — the pillars. These are the scenes where the protagonist's situation changes irreversibly, where the central question is asked or answered, where the emotional cost is highest. If I removed a pillar, the book would collapse.
Then I ask: does every non-pillar scene either build toward a pillar, pay off a pillar, or connect two pillars? If a scene does none of those things, it is either decoration or dead weight. Decoration has a place — novels need breathing room, and a scene that exists purely for texture or pleasure can earn its keep. But it should be a choice, not an accident.
The pillar test is also how I decide restructuring. If the pillars are in the right order and the right places, the structure is probably sound even if the connective tissue is messy. If the pillars themselves are in the wrong order — if the emotional climax comes before the plot climax, or if the revelation that should anchor the middle arrives too early — then restructuring is necessary, and everything else is secondary.
The Emotional Arc vs. the Plot Arc
One of the structural edit's most important jobs is checking whether the emotional arc and the plot arc are aligned. They often are not, in a first draft, because the writer was focused on getting the plot from A to B and let the emotional development happen wherever it happened.
The plot arc is what happens. The emotional arc is what the protagonist feels about what happens. A novel can have a strong plot arc and a broken emotional arc: the events are coherent, but the protagonist's inner journey jumps from grief to joy to grief again without the reader seeing the transitions. Or it can have a strong emotional arc and a weak plot arc: the protagonist's feelings are rich and continuous, but nothing in the external world puts pressure on them.
In the structural map, I track both. For each scene: what changes in the world? What changes in the protagonist? If the world changes but the protagonist does not, the scene is plot without consequence. If the protagonist changes but the world does not, the scene is interior without stakes. The best scenes change both. A novel that has only one kind will feel thin no matter how good the prose is.
AI can help with this audit by asking the model to track the protagonist's emotional state chapter by chapter. The model will produce a simplified, sometimes reductive emotional timeline. That is fine. You are not looking for nuance. You are looking for discontinuities — places where the emotional state jumps without a cause, or where it stays flat for too long. Those are structural problems, not line-level problems.
What the Structural Edit Is Not
The structural edit is not a polish. I do not rewrite sentences. I do not improve metaphors. I do not fix dialogue. If I notice a bad sentence, I put a checkmark in the margin and keep reading. The checkmark means "come back in the line edit." If I stop to fix the sentence, I lose the cold reader's perspective. The whole method depends on maintaining distance. Polishing pulls you back inside.
The structural edit is also not a copyedit. I do not check consistency of names, timelines, or facts. That comes later. I am aware of inconsistencies when I see them — the character's eye color that changed, the timeline that has two Tuesdays in a week — but I note them and move on. They are real problems, but they are not structural problems, and fixing them now would pull me into the page when I need to be above it.
Finally, the structural edit is not a rewrite. There is a moment in many writers' lives where they read their first draft and conclude that the whole thing is worthless and they need to start over. This feeling is almost always wrong. The draft is not worthless. It is unfinished. The structural edit is how you finish it. Starting over is how you spend five years on the same novel without ever holding a completed one.
The Shape After the Edit
When the structural edit is done, the book should feel different. Not polished — that comes later — but coherent. The scenes that remain all have a reason. The order makes sense. The pillars are in place. The emotional arc tracks with the plot arc. The promises are paid. The dead weight is gone.
You will also have a list: scenes to write, scenes to move, a restructuring plan if needed, and a margin full of checkmarks for the line edit. That list is the map for the next two passes: the scene deepening pass (Wednesday) and the line edit (Friday). The structural edit produces the plan. The next passes execute it.
This is the unglamorous truth about novels. The first draft gets the glory because it is where the story is discovered. The line edit gets the satisfaction because it is where the prose becomes beautiful. The structural edit gets neither. It is hard, slow, unphotogenic work. You are not writing. You are not polishing. You are reading your own book as if it were someone else's and deciding what it needs to become itself.
But this is the pass that separates a draft from a novel. Do it honestly, and every pass after it becomes easier. Skip it, and you will spend the rest of your revision polishing a structure that cannot hold.
For Next Time
On Wednesday we move from architecture to flesh: ["Novel I — Revision: Scene Deepening"](/harrys-desk/novel-i-revision-scene-deepening). The structural edit tells you which scenes belong. The scene-deepening pass tells you whether they are doing enough. We will talk about the layers of a scene — stakes, subtext, sensory ground, and the gap between what a character says and what a reader understands — and how to use AI to pressure-test a scene without letting it flatten the tension.
Your homework until then: take the structural map of your novel — or, if you have not finished a draft, take the last three chapters of any novel you admire — and identify the pillars. Which scenes are load-bearing? Which scenes are decoration? Which scenes are dead weight? Then ask the hardest question: if you removed any pillar, would the book still stand? The answer tells you whether your structure is holding weight or just filling space.
---
*Harry Mercury, Editor in Chief* *The SMF Works Project* *Week 13, Article 1*