Editing vs. Generating: Why the Prompt Is Different
Most guides to AI prompting are written for generation: "Write a story about..." or "Create a scene where..." But if you're a fiction writer with a manuscript you've already written, generation prompts are the wrong tool. You don't want AI to write your book. You want it to help you make your book better.
Editing prompts require a fundamentally different approach. Instead of giving the AI creative latitude, you're giving it constraints. Instead of asking "what would you write," you're asking "what's wrong with what I wrote, and how would you fix it while keeping my voice." This distinction sounds subtle, but it determines whether you get back something that sounds like you or something that sounds like AI.
The Anatomy of a Good Editing Prompt
An effective fiction editing prompt has three components:
- Context: What the AI needs to know about the text it's editing
- Instruction: What specific edit you want made
- Constraints: What should NOT change
Most writers include the first two and forget the third. Constraints are what keep AI from rewriting your entire paragraph when you only wanted to fix one sentence.
A Basic Structure
Here's a prompt template that works well for most fiction editing tasks:
"Here is [a paragraph / a scene / a chapter] from my novel. [Brief context about the story if needed.] I want you to [specific editing instruction]. Keep my voice and style intact -- don't add flowery language or change the tone. Only modify what's necessary to achieve the edit I described."
That last sentence does an enormous amount of work. Without it, most AI models will "improve" things you didn't ask them to touch. For a deeper exploration of why this constraint matters, see our guide on how to use AI to edit fiction without losing your voice.
Specific Prompt Structures by Editing Task
Tightening Prose
Loose prose is one of the most common first-draft issues, and AI handles it well because the task is relatively objective: remove redundancy, eliminate filler, strengthen verbs.
Good prompt: "Tighten this paragraph. Cut unnecessary words, replace weak verbs with stronger ones, and eliminate redundant phrases. Don't change the meaning, the character's voice, or the information conveyed. Only tighten."
Bad prompt: "Make this better."
The bad prompt gives the AI permission to change anything, and it will. "Make this better" is an invitation to rewrite. "Tighten this paragraph" is a scalpel.
Fixing Pacing
Pacing prompts need to specify direction. "Fix the pacing" is too vague -- does it move too fast or too slow?
Good prompt: "This scene moves too quickly. The reader needs more time to feel the weight of what just happened. Expand the moment after Sarah receives the news -- add interiority, sensory detail, or a small physical action that slows the reader down. Don't add more than 2-3 sentences."
Good prompt (opposite direction): "This passage drags. The reader already knows the characters are nervous -- the three paragraphs describing nervous behaviors are redundant. Cut it down to one paragraph that conveys the same anxiety with fewer words."
Notice how both prompts are specific about what's wrong and what the fix should look like. They also include a constraint on scope -- "don't add more than 2-3 sentences" or "cut it down to one paragraph."
Strengthening Dialogue
Dialogue editing is where many writers first discover AI's usefulness, because hearing the rhythm of dialogue on the page is genuinely difficult.
Good prompt: "This dialogue between Marcus and Elena sounds too similar -- they both speak in the same rhythm and vocabulary. Marcus is a working-class mechanic who didn't finish high school. Elena is an academic who overexplains things. Revise their lines so each character sounds distinct. Keep the content of what they say the same."
Bad prompt: "Make this dialogue more realistic."
The good prompt tells the AI who these characters are and what makes them different. The bad prompt leaves the AI to guess, and its guess will be generic.
Show Don't Tell
Good prompt: "This sentence tells the reader that James was angry. Rewrite it to show his anger through a physical action, a dialogue choice, or an environmental observation. Don't use the word 'angry' or any synonym for it."
Bad prompt: "Show don't tell this."
The good prompt specifies what "showing" means for this particular instance. The bad prompt assumes the AI shares your understanding of the principle and will apply it the way you want.
Voice and Tone Adjustments
Good prompt: "This passage is written in a neutral, distant third person. I need it to feel more intimate -- closer to the character's thoughts. Shift toward a close third-person perspective where the narration reflects how the character perceives the world. Don't switch to first person."
Bad prompt: "Make this more emotional."
"More emotional" could mean melodramatic, sentimental, raw, or a dozen other things. Specifying the mechanism (close third person, interiority) gives the AI a technique to apply rather than a vague direction to interpret.
Anti-Patterns: Prompts That Damage Your Writing
"Make This Better"
This is the single most common -- and most damaging -- prompt writers use. It gives the AI maximum latitude to change anything, which means it will impose its own defaults on your prose. AI defaults tend toward a certain kind of polished, MFA-workshop prose that sounds competent but generic. If you want your writing to sound like AI-edited writing, "make this better" will get you there.
"Rewrite This Scene"
Unless you genuinely want a rewrite from scratch, this prompt destroys your voice. The output will be the AI's scene, not yours. If you want to keep your scene but fix specific problems, name those problems.
"Add More Description"
Without specifying what kind of description, you'll get generic sensory details that don't serve the story. Better: "Add a visual detail in this paragraph that reinforces how trapped the character feels." Give the description a purpose.
"Fix the Grammar"
In fiction, "grammar" is complicated. First-person narrators have distinctive grammatical patterns. Dialogue should have grammatical errors that match the character's speech. Stream of consciousness breaks rules intentionally. If you ask AI to "fix the grammar," it may standardize your intentional choices into bland correctness. Better: "Fix any unintentional grammatical errors. Preserve any non-standard grammar that serves a stylistic or characterization purpose."
Not Specifying What Should Stay
This is the meta anti-pattern. Any prompt that doesn't tell the AI what to preserve is a prompt that risks changing things you didn't want changed. Explicit constraints ("keep the metaphor about the river," "don't change the last sentence," "maintain the fragmented sentence structure") protect the elements of your writing that are working.
Advanced Techniques
The Two-Pass Prompt
For complex edits, break them into passes. First ask the AI to identify problems, then ask it to fix them. This prevents the AI from making changes you don't agree with.
Pass 1: "Read this chapter and identify the three biggest pacing issues. Don't edit anything yet -- just tell me what you see and where."
Pass 2: "Good, I agree with issues 1 and 3 but not 2. Please fix issue 1 by [specific instruction] and issue 3 by [specific instruction]. Leave everything else unchanged."
This gives you editorial control over what gets changed rather than accepting a bundle of edits you didn't all ask for.
The "Edit Like" Prompt
If you have a specific editorial sensibility in mind, reference it:
"Edit this passage the way a developmental editor would -- focus on whether the scene achieves its narrative purpose, not on line-level prose. Is the character's motivation clear? Does the conflict escalate? Does the scene end at the right moment?"
Or: "Edit this passage the way a line editor would -- focus on sentence-level clarity, rhythm, and word choice. Don't change the structure or content."
Assigning a role constrains the type of feedback and keeps the AI from mixing developmental and line editing in ways that confuse the revision process.
The Comparison Prompt
When you're not sure which direction to take, ask for options rather than a single edit:
"Give me three different ways to rewrite this opening sentence. The first should be shorter and punchier. The second should set the scene with a sensory detail. The third should start with dialogue. Keep the same information in all three."
This turns the AI into a brainstorming partner rather than a decision-maker. You choose which version works for your story.
Voice Calibration
If you're going to use AI for multiple editing sessions, spend time upfront calibrating it to your voice:
"Here are three paragraphs from my novel that I'm happy with. Study the sentence length patterns, vocabulary level, use of figurative language, and tone. When you edit other passages from this manuscript, match this style. Don't elevate the vocabulary, don't make sentences more complex, and don't add metaphors unless the existing text already uses them."
This is the single most important prompt for writers who use AI regularly. It's the difference between AI that edits like your ideal editor and AI that edits like a generic writing instructor.
The Voice Editing Advantage
One reason many writers find voice-based editing interfaces (like Fable's voice editing feature) more natural than typing prompts is that speaking your editing intention is inherently more specific. When you say "that dialogue feels stilted -- make Marcus sound more like he's avoiding the real topic," you're giving context, direction, and constraint in natural speech without having to think about prompt engineering.
Typed prompts tend to be either too short ("fix this") or overthought ("please analyze the narrative discourse structure and optimize the pragmatic implicature"). Spoken editing instructions land in the middle, where the best prompts live. For a practical walkthrough of this approach, see our guide on how to edit a novel with voice commands.
The Golden Rule of AI Editing Prompts
Here it is: the specificity of your prompt should match the specificity of the edit you want.
If you want a small change (tighten one sentence), give a narrow prompt with tight constraints. If you want a bigger change (restructure a scene), give a broader prompt but still specify what should be preserved. Never give a broad prompt for a narrow problem or a narrow prompt for a broad one.
AI models are extraordinarily capable editors when directed well. The bottleneck is almost always in the prompting, not the model. A mediocre model with a great prompt will outperform a great model with a vague prompt every time -- and model choice still matters, so see our Claude vs. ChatGPT comparison for creative writing if you are choosing between platforms. Learn to prompt for editing, not generation, and AI becomes the tireless, infinitely patient revision partner that every writer needs but few can afford to hire.