Outlines, character work, dialogue passes and getting unstuck. Below are 7 copy-ready prompts. Fill in the [BRACKETS], copy, and paste into ChatGPT, Claude, Gemini or any capable assistant.
Fiction is where AI is most useful as a tool and least useful as a writer. Everything structural it does well; everything that makes a reader feel something remains yours.
The 7 prompts
Develop a character beyond a description
Build a character with contradictions rather than a list of traits.
Help me develop this character. WHAT I HAVE SO FAR: [NAME, ROLE IN THE STORY, ANY TRAITS] STORY CONTEXT: [GENRE, SETTING, WHAT THE STORY IS ABOUT] THEIR FUNCTION IN THE PLOT: [WHAT THEY DO FOR THE STORY] Produce: 1. WHAT THEY WANT - the stated goal, and the deeper want underneath it that they may not articulate. These should not be the same thing. 2. WHAT THEY FEAR - and how the fear and the want conflict. A character who wants something with no cost is not a character. 3. THE CONTRADICTION - the trait that does not fit the others. Consistent characters are flat; the interesting thing is the generous person who is cruel about one specific thing, or the coward who is fearless about one thing. 4. THEIR BLIND SPOT - what they are wrong about themselves, and how the reader will see it before they do. 5. HOW THEY SPEAK - vocabulary, sentence length, what they avoid saying, what they say when uncomfortable, and one verbal habit. Give three sample lines in three different situations: relaxed, under pressure, lying. 6. THE PHYSICAL - two or three specific details that reveal character rather than appearance. What they do with their hands, how they occupy a room, what they wear that is wrong for the occasion. Avoid hair and eye colour; they reveal nothing. 7. WHAT THEY WOULD NEVER DO - the line they will not cross. This defines a character more sharply than anything they would do, and it is where the plot can apply pressure. 8. THE SCENE THAT REVEALS THEM - a specific situation where all of the above would show without being explained. 9. HOW THEY CHANGE - or pointedly do not, and what the story does to cause or refuse it. Avoid backstory for its own sake. Only include history that shapes present behaviour.
What you get: A character built on conflicting wants, a contradiction, a blind spot and a line they will not cross, with sample dialogue.
Tip: Point 3 is where characters come alive. Perfectly consistent characters read as arguments rather than people.
Diagnose why a scene is not working
Fix a scene that feels flat without knowing why.
This scene is not working. Diagnose it. SCENE: """ [PASTE] """ WHAT IT IS MEANT TO DO IN THE STORY: [PURPOSE] WHAT COMES BEFORE AND AFTER: [CONTEXT] WHAT FEELS WRONG TO ME: [IF YOU KNOW] Check against the common causes, in this order: 1. NOBODY WANTS ANYTHING - the most common cause. What does each character want in this scene, specifically, right now? If a character has no want, they are furniture. 2. NO OBSTACLE - the want is achieved too easily, or nothing opposes it. 3. NOTHING CHANGES - what is different at the end of the scene from the start? If the answer is 'the reader knows more', it is an information delivery, not a scene. 4. IT STARTS TOO EARLY - where could this scene begin instead? Almost always later than it does. 5. IT ENDS TOO LATE - where could it end? Usually before the current ending, on the moment of change rather than after the aftermath. 6. THE WRONG POV - would this be better from another character's perspective? Usually the scene belongs to whoever has most to lose. 7. TOLD NOT SHOWN, OR SHOWN WHEN IT SHOULD BE TOLD - both are errors. Quote examples of each if present. Not everything needs dramatising; summary has its place. 8. ON-THE-NOSE DIALOGUE - characters saying exactly what they mean and feel. 9. NO SENSORY ANCHOR - the scene happens nowhere in particular. For each: does it apply? Quote the evidence. Then: - THE DIAGNOSIS - the single main problem - THE FIX - what to change, specifically - THE HARDER QUESTION - does this scene need to exist? If its purpose could be served in two lines elsewhere, say so.
What you get: A cause-by-cause diagnosis with quoted evidence, a single main problem, and a check on whether the scene should exist.
Tip: Point 1 is the answer roughly half the time. A scene where nobody wants anything cannot be fixed with better prose.
Structure a story that keeps stalling
Find the shape of a story you have been circling.
Help me structure this story. WHAT I HAVE: [PREMISE, CHARACTERS, SCENES, FRAGMENTS - whatever exists] LENGTH: [short story / novella / novel] GENRE AND TONE: [DETAIL] WHERE I AM STUCK: [DESCRIPTION] THE ENDING, IF I KNOW IT: [OR 'no idea'] Produce: 1. WHAT THE STORY IS ABOUT - two answers. The plot (what happens) and the theme (what it is really about). If I cannot tell you the second, the structure will not hold, and that is usually the reason for being stuck. 2. THE CENTRAL QUESTION - the dramatic question the reader is waiting to have answered. Phrase it as a yes/no question about the protagonist. 3. THE PROTAGONIST'S WANT AND NEED - what they are chasing, and what they actually require, which is usually different. The gap between them is the story. 4. THE STRUCTURE - beats appropriate to my stated length, each with: what happens, what changes, and which way the central question tips. Do not impose a template; derive it from my material and say which shape it wants. 5. THE MIDDLE PROBLEM - if I am stuck, it is probably the middle. Diagnose: is the obstacle not escalating, is the protagonist passive, or is there no ticking clock? Name which and fix it. 6. WHAT IS MISSING from what I gave you: a turn, a cost, a reversal, a reason the protagonist cannot simply walk away. 7. THE ENDING - if I do not know it, work backwards from the central question and give three possible answers, each with what it would mean thematically. 8. WHAT TO CUT - material I have that does not serve this structure. Say so plainly; being stuck is often the result of protecting a scene that does not belong. Do not write the story. Structure it.
What you get: Plot and theme separated, a central dramatic question, a derived structure, and a diagnosis of the middle.
Tip: Point 1's theme question is usually where stuck stories unstick. Not knowing what the story is about makes every structural choice arbitrary.
Get critique on a piece of fiction
Have your writing read honestly and usefully.
Critique this fiction. Be honest - encouragement is not useful to me. TEXT: """ [PASTE] """ GENRE AND INTENDED AUDIENCE: [DETAIL] WHAT STAGE IT IS AT: [first draft / revised / near final] WHAT I WAS TRYING TO DO: [INTENT] WHAT I AM WORRIED ABOUT: [YOUR CONCERN] Produce: 1. WHAT IT IS DOING WELL - specific, quoted. Not encouragement; identification. If a technique is working, name the technique. 2. THE READER EXPERIENCE - describe honestly what it is like to read this: where attention was held, where it drifted, where you were confused, where you were ahead of the text. This is more useful than a verdict. 3. DID IT DO WHAT I INTENDED - compare the effect to my stated intent. Where they diverge is the finding. 4. CRAFT NOTES BY LEVEL: - STORY: stakes, causality, momentum, whether anything changes - CHARACTER: whether they read as people, whether their choices follow from who they are - SCENE: where scenes start and end, what each accomplishes - PROSE: rhythm, clarity, over-writing, tics. Patterns rather than every instance. 5. THE ONE THING - if I revise one thing, what. 6. THE THING I AM AVOIDING - if there is a problem the piece works around rather than solves, name it. 7. WHAT IS TASTE AND WHAT IS CRAFT - separate your preferences from actual problems. Be explicit about which is which. 8. THE HONEST ASSESSMENT - given the stage I said it is at, where does it stand? Do not inflate. Rules: - Quote what you are talking about - Do not rewrite my prose. Diagnose. - Do not soften the main problem by burying it among minor ones
What you get: A reader-experience account, level-separated craft notes, one priority revision, and taste separated from craft.
Tip: Point 2 beats a verdict. 'I drifted at the second paragraph of page three' is actionable; 'the pacing is uneven' is not.
Build a setting that does work in the story
Develop a world that affects the plot rather than decorating it.
Develop this setting. WHAT I HAVE: [PLACE, TIME, ANY DETAILS] STORY: [WHAT HAPPENS THERE, WHO THE CHARACTERS ARE] GENRE: [DETAIL] HOW MUCH SETTING MATTERS: [backdrop / significant / the setting is the point] Produce: 1. THE ONE RULE - if this is a speculative or secondary world, the single premise everything else follows from. Worlds built from one rule cohere; worlds built from a list of features do not. 2. HOW THE SETTING CONSTRAINS THE PLOT - what it makes hard, impossible, or costly for my characters. A setting that does not constrain anyone is wallpaper. 3. THE SENSORY SPECIFICS - five concrete details, chosen for what they reveal rather than for atmosphere. Include at least one sound and one smell; writers default to the visual. 4. THE DAILY TEXTURE - what ordinary life is like here: how people earn, eat, travel, argue, and what they complain about. Mundane detail convinces readers far more than grand description. 5. WHO HAS POWER AND WHY - and how an ordinary person experiences that. 6. THE CONTRADICTION - something about this place that does not fit the obvious picture. Real places are inconsistent. 7. WHAT THE CHARACTERS DO NOT NOTICE - the things natives take for granted, which the prose should therefore not explain. This is the main technique for avoiding exposition. 8. HOW TO REVEAL IT - three moments in the story where setting could be conveyed through action rather than description. 9. WHAT TO LEAVE OUT - what I do not need to work out. Over-built worlds leak into the text as exposition. Name the areas to leave vague. Do not produce an encyclopaedia. Produce what the story can use.
What you get: A setting derived from one rule, with plot constraints, daily texture, a contradiction and an explicit leave-vague list.
Tip: Point 9 is the discipline that keeps worldbuilding from swallowing the book. The detail you invent wants to get onto the page whether it belongs there or not.
Fix pacing across a whole draft
Work out why a manuscript drags or rushes.
Analyse the pacing of this draft. DRAFT OR CHAPTER SUMMARIES: """ [PASTE - FULL TEXT IF SHORT, OR A SUMMARY OF EACH SCENE/CHAPTER WITH WORD COUNTS] """ LENGTH AND GENRE: [DETAIL] WHERE READERS SAID IT DRAGS: [FEEDBACK, IF ANY] Produce: 1. THE SCENE MAP - a table: Scene | Words | % of total | What changes | Tension (rising/flat/falling) | Type (action / dialogue / interiority / description / transition) 2. THE FLAT STRETCHES - runs of scenes where nothing changes or the same thing changes repeatedly. Repetition of function is the most common pacing fault: three scenes that all establish the same thing feel slow even when each is well written. 3. PROPORTION - where word count does not match story importance. Long treatments of minor events and quick treatments of major ones. 4. RHYTHM - the sequence of scene types. Too many consecutive scenes of the same type flattens pace regardless of content. Four dialogue scenes in a row drags even if each is good. 5. THE MIDDLE - specifically assess the middle third. Is the protagonist driving events or reacting? Passivity is the usual cause of a sagging middle. 6. TENSION CURVE - describe the shape across the draft. Where does it plateau, and for how long? 7. WHAT TO CUT - specific scenes, ranked. For each: what is lost and where its necessary function could be served instead. 8. WHAT TO EXPAND - the rushed parts. Usually the emotional aftermath of major events, which writers skip past. 9. THE THREE CHANGES that would most improve pace, in order. Be specific about scenes. 'The middle drags' is not a finding.
What you get: A scene-by-scene map with tension and type, flat stretches identified, and a ranked cut and expand list.
Tip: Point 4 explains pacing problems that survive every other fix. Well-written scenes of the same type stacked together read as slow no matter how good each one is.
Get unstuck when you do not know what happens next
Break a block by generating options rather than waiting.
I am stuck. Help me find what happens next. THE STORY SO FAR: [SUMMARY] WHERE I AM STUCK: [THE EXACT POINT] WHAT I HAVE TRIED: [ANY IDEAS YOU REJECTED AND WHY] WHAT I KNOW ABOUT THE ENDING: [OR 'nothing'] CHARACTERS AND WHAT THEY WANT: [DETAIL] First diagnose - being stuck usually has one of these causes: - The protagonist has no reason to continue. Their want has been satisfied or abandoned. - The obstacle is not escalating; the next scene would repeat the last. - You have written the character into someone who would not do what the plot needs. - You do not know what the story is about, so you cannot tell which option is right. - The problem is earlier. The block is a symptom; something three chapters back closed off the path. Say which applies, with evidence from what I told you. Then produce: 1. TEN THINGS THAT COULD HAPPEN NEXT - genuinely varied. Include: the obvious one, the opposite of the obvious one, one where a character refuses to do what the plot wants, one where something is revealed to have been untrue, one where a minor character acts, one where the protagonist fails, and one where the situation gets worse for a reason unrelated to the current conflict. 2. FOR THE BEST THREE - what each would mean for the rest of the story, and which best serves what the story seems to be about. 3. THE QUESTION TO ANSWER - what I need to decide before this scene can be written. Often the block is an undecided question rather than a missing idea. 4. IF THE PROBLEM IS EARLIER - where, and what would need to change. 5. THE PERMISSION - one thing I appear to be protecting that could be cut or broken. Blocks often come from defending a scene, a character, or a plan that no longer fits. Do not write the scene. Give me options and a diagnosis.
What you get: A diagnosis of why you are stuck, ten genuinely varied options, and the undecided question behind the block.
Tip: Point 4 is right more often than it feels. A block at chapter twelve is frequently a wrong turn at chapter eight finally becoming visible.
Where AI actually helps here
- Breaking a stuck scene by generating ten wrong versions
- Continuity and consistency across a long manuscript
- Interrogating a character — ask it to play them and ask hard questions
Where it falls down
- Prose. Generated prose is competent and forgettable, and the rhythm gives it away
- Subtext. It states what should be implied, every time
- Endings. It reaches for resolution and closure by default
The mistake almost everyone makes: Asking it to write the scene
Ask it to describe what has to happen in the scene, what each character wants, and what the reader should feel by the end. Then write the scene yourself. The planning is genuinely helpful; the prose is the part you are writing for.
Free tool: Prompt Chain Builder
Runs in your browser. No sign-up, nothing uploaded.
Questions people ask
Can AI write a novel?
It can produce novel-length text. Whether that is a novel is a different question — and the practical problem is that it cannot hold a through-line across 90,000 words, so the middle drifts.
Do I have to disclose AI use in fiction?
It depends entirely on where you are submitting. Many publishers, magazines and competitions now require disclosure or prohibit it outright, and terms differ. Check the specific guidelines — this is not a general rule.