foundations · Level 2
Writing fiction with AI: what actually helps
The honest, evidence-based split: where AI genuinely helps a novelist (brainstorming, unsticking a scene) and where most working novelists still do not trust it (voice, structure, the whole draft).
Your next step
Start this course online
Register your start, then work through the lessons and exercises at your own pace. You can download the PDF separately. Where a scored exam is available, you will find it after the lessons.
Register or resume your course
Start with the essentials
The short answer
The evidence on AI and fiction writing genuinely disagrees. One November 2025 survey found 61% of working writers use AI at least sometimes, yet within that same study only 11% of the 291 fiction authors surveyed used it for text they actually publish. AI reliably helps with brainstorming and unsticking a scene. Most novelists still do not trust it with voice, structure or a full draft.
What you will learn
- You will be able to explain which fiction-writing tasks AI genuinely helps with, and which most novelists still do not trust it for.
- You will understand why the evidence on fiction-writer AI adoption is contested, and be able to explain the difference between using AI sometimes and publishing AI-assisted text.
- You will be able to write a brainstorming prompt that gets you unstuck on a scene, rather than asking AI to write the scene for you.
- You will know what a manuscript-editing tool can and cannot judge about your craft.
- You will have a short, honest, written policy for which parts of your writing process you will and will not hand to AI.
Who it is for
Fiction writers and novelists, hobbyist or working, who want an honest, evidence-based answer to where AI actually helps their writing practice and where it does not, rather than a hype pitch or a blanket dismissal.
Before you start
- None. It helps to have read Your first prompts first if you have never written a prompt before.
Learn in your browser · Chapter 01 of 06
The honest evidence: what the numbers on AI and fiction writing actually say
Before any technique, get the numbers straight, because they genuinely disagree, and understanding why they disagree is the first real lesson in this workbook.
A survey of 1,481 working writers, including 291 fiction authors, found that 61% of all respondents said they use AI at least sometimes. That headline figure is the one that tends to travel furthest. Look only at the fiction authors inside that same survey, though, and the figure drops to 42%, and only 11% of them said they use AI to create text they actually publish.
A separate survey of 258 published UK novelists, run by the University of Cambridge's Minderoo Centre for Technology and Democracy and published on 20 November 2025, found a similar shape from a different angle: 33% of novelists use AI somewhere in their writing process, mostly for non-creative tasks like sourcing facts, and only around 8% use it for editing text. The same survey found 97% of novelists extremely negative about the idea of AI writing a whole novel, and 87% extremely negative about it writing even a short section.
A worked example (invented). Two novelists are asked, separately, whether they use AI. The first says yes, meaning she once asked a chatbot for ten alternative titles and used none of them. The second says no, meaning he has never let anything but his own hands touch a sentence he intends to publish, even though he used an AI tool last week to check when a specific model of car was first sold. Both answers are honest. Neither tells you much on its own, which is exactly why this workbook asks a narrower, more useful question chapter by chapter: which specific task, not whether AI is used at all.
Try it yourself · Activity 01
8 minAnswer the narrower question about yourself
Before learning any technique, get honest about your own current AI use as a writer, using the same three-way split the surveys use.
- Write down whether you have ever experimented with AI for any part of your writing, even once, out of curiosity.
- Write down whether you use it at least sometimes as part of your regular process.
- Write down whether any text you have actually published or shared with readers contains a sentence AI wrote that you kept largely as it was written.
Which of the three answers surprised you when you wrote it down honestly, and why?
Learn in your browser · Chapter 02 of 06
Where AI earns its keep: brainstorming and getting unstuck
The single most consistently reported use across the surveys in this workbook is not writing prose. It is helping a writer think through options faster than staring at a blank page alone.
Brainstorming, in this context, means using AI to generate a wide spread of options, character motivations, plot directions, opening lines, so that you can choose or reject them, rather than asking it to write the finished version of any of them. This is the task every survey in this workbook finds novelists reaching for first, and it is also the task where a weak prompt fails hardest.
A worked example. A weak prompt: 'Give me ideas for what happens next in my heist novel.' This returns generic caper tropes with no connection to your actual characters or stakes, because the AI has nothing specific to work from. A rebuilt, stronger prompt: 'My character Elena has just realised her partner tipped off the target. She has thirty seconds before he notices she knows. Give me five different ways she could react in the moment, ranging from confronting him outright to pretending she has not noticed, and for each one, name the specific complication it creates for the rest of the heist.' The second prompt gives the model your actual stakes, your actual character, and a constraint (five options, each with a stated consequence), which turns a vague request into a genuinely useful shortlist.
- A brainstorming problem: you know roughly what needs to happen next but cannot see the shape of it, or every direction you try feels obvious.
- A voice problem, not a brainstorming problem: you know exactly what happens next, but every sentence you write sounds flat or wrong. AI brainstorming will not fix this. Chapter 4 covers why.
- A research problem, not a brainstorming problem: you are stuck because you do not know a fact you need (a procedure, a place, a date). Chapter 3 covers this instead.
Try it yourself · Activity 02
15 minGet unstuck on a real scene
Pick a scene in your own work in progress that you are currently stuck on, or invent a simple one if you are between projects.
- Write one sentence stating exactly where the scene is stuck and what the character wants at that moment.
- Write a brainstorming prompt following the worked example above: your character's situation, the constraint they are under, and a request for five distinct directions, each with a named consequence.
- Run the prompt in whichever AI tool you already use, or draft the five directions yourself by hand if you would rather not use a tool for this exercise.
- Pick the one direction that surprised you most, and write two sentences on why it works better than what you had before.
Did the useful part come from an idea AI generated outright, or from something it prompted you to think of yourself?
Learn in your browser · Chapter 03 of 06
Research and getting the facts right, without losing your reader's trust
When novelists in the Cambridge survey said they used AI, the most common reason was not creative at all: sourcing facts and information for a scene.
Around 20% of the novelists in that survey use AI specifically to source facts, a period detail for a historical scene, a technical process for a thriller, the correct term for a piece of equipment. This is a sensible use of a fast, conversational search tool, provided you treat what comes back as a first pass, not a finished fact.
Scroll sideways to see every column.
| What you asked AI to source | Verify it against |
|---|---|
| A period detail (what people ate, wore or said in a given decade) | A dedicated history site, museum resource or a book on the period, not a single chatbot answer |
| A technical or medical procedure your plot depends on | A primary professional source, or someone who actually does the job, if the detail is load-bearing for the plot |
| A place name, distance or travel time | A map or an official source for the place itself |
| A law, regulation or court process | The relevant official body's own published guidance, never AI's own summary of it |
Try it yourself · Activity 03
12 minSource a fact, then check it yourself
Pick one factual detail your current work in progress genuinely needs, something you have been guessing at or have not looked up yet.
- Ask an AI tool to source the fact for you, and note exactly what it told you.
- Independently verify the same fact using a source from the table above, not another AI tool.
- Write down whether the two answers matched, and if they did not, what the AI got wrong or oversimplified.
If you had put the AI's first answer straight into your manuscript without checking, what would a well-informed reader have caught?
Learn in your browser · Chapter 04 of 06
Where most novelists still do not trust it: voice, structure and the whole draft
The same Cambridge survey found overwhelming resistance to AI touching a whole novel, and only slightly less resistance to a short section. This chapter looks at why voice and structure are a different kind of problem from brainstorming or fact-checking.
97% of the novelists surveyed were extremely negative about the idea of AI writing an entire novel, and 87% extremely negative about it writing even a short section, with resistance easing only for editing, where 43% were still extremely negative. A broader August 2026 survey of working writers generally, not fiction specific, found a related pattern from the inside: 71% had used AI, yet 60% of all the writers surveyed said they distrusted AI-supported writing, including work they had produced themselves.
Voice is the sum of thousands of small, specific choices, a writer's particular rhythm, the words they reach for under pressure, the jokes only their narrator would make, held consistently across an entire book. A model trained to predict the statistically likely next word tends to smooth toward an average, competent, slightly generic middle, which is precisely where a distinctive voice needs to be sharp rather than average.
Structure, across a whole manuscript, means holding a throughline in your head: a detail planted in chapter three that pays off in chapter twenty, a subplot's pacing relative to the main plot, a character arc that has to land on the right beat at the right moment. None of the surveyed uses in this workbook, brainstorming a single scene, sourcing a single fact, show novelists trusting AI with that kind of whole-manuscript judgement, and the resistance figures above suggest most are not close to changing their mind on the whole draft.
Try it yourself · Activity 04
12 minMake a line yours again
Take a competent but generic sentence, either one AI has actually suggested to you, or the invented example above if you have nothing to hand.
- Rewrite the sentence in your own voice, changing whatever needs to change so it sounds like your narrator rather than a generic one.
- List three specific things you changed: a word choice, a rhythm, a detail only your character would notice.
- Decide whether any of your three changes could have come from an AI brainstorming prompt, or whether all three needed to come from you.
What does this exercise suggest about the actual boundary between AI as a thinking partner and AI as the writer?
Learn in your browser · Chapter 05 of 06
Editing tools and the disclosure question
Beyond brainstorming and research sits a narrower, genuinely useful category: tools built specifically for manuscript craft, and a newer question that comes with using any of them if you plan to publish.
Two real, named tools illustrate the category. ProWritingAid is built specifically for long-form fiction manuscripts, checking things a general grammar checker is not tuned for: overused words across an entire book, pacing flags, dialogue tag habits. Sudowrite is built for fiction drafting and brainstorming specifically, rather than general-purpose writing. Naming them here is factual, not an endorsement or a partnership: they are simply real, current examples of tools purpose-built for this category, and better options may exist by the time you read this.
- What a manuscript-editing tool tends to catch well: repeated words across chapters, timeline inconsistencies (a character's eye colour changing), overlong sentences, pacing flags at a chapter level.
- What it still cannot judge: whether a plot twist actually lands emotionally, whether a character's choice feels true to who they are, whether your risk in breaking a grammar rule on purpose is working.
Try it yourself · Activity 05
15 minRun a page through the checklist
Take one page of your own manuscript, ideally one you have not recently reread.
- Run it through a grammar or manuscript-consistency tool if you have one, or work through the list above by hand if you do not.
- Note every genuine catch: something you had actually missed.
- Note one flag that was wrong, a deliberate choice the tool mistook for an error.
- Decide, in one sentence, what this exercise tells you about how much editing judgement you are willing to hand over.
Was the tool more useful for catching small, mechanical errors or for judging anything about the story itself?
Worked answer
Manuscript-editing tools are consistently strongest at the small, mechanical layer: repeated words, timeline slips, overlong sentences, pacing flags. They have no reliable way to judge whether a twist lands emotionally or a character's choice feels true, because that judgement depends on the whole story's context, which is exactly the kind of whole-manuscript reasoning chapter 4 found novelists trusting AI with least.
Learn in your browser · Chapter 06 of 06
Your honest AI-writing policy, and the final exercise
Everything so far comes together into a short, personal policy you can actually hold yourself to, then this workbook's closing page.
- Use AI for brainstorming and getting unstuck: generating options you choose from, not finished prose you keep as written.
- Use it to source facts fast, then verify anything you would be embarrassed to get wrong, from a source you trust, before it reaches the page.
- Do not let it write your voice: if a sentence sounds like it could belong to anyone, rewrite it until it can only belong to your narrator.
- Use a purpose-built editing tool for the mechanical layer, and keep your own judgement for whether the story actually works.
- Check the specific AI-disclosure rules for wherever you plan to publish before you submit.
- Revisit this policy every few months. The tools and the evidence behind this workbook will both keep moving.
Try it yourself · Activity 06
10 minWrite your own policy
Using the six points above as a starting frame, write your own short, honest policy in your own words.
- Write four to six lines stating exactly what you will use AI for in your writing practice.
- Write one line stating what you will never hand to it, in your own words, not a copy of this workbook's.
- Put the policy somewhere you will actually see it next time you open your work in progress.
Which single line of your policy would you be most tempted to break under deadline pressure, and why?
What you can do now. You know the real, contested evidence on how many fiction writers use AI, and why the numbers genuinely disagree rather than one of them being wrong. You have practised the one task every survey in this workbook agrees AI is actually good for, brainstorming and getting unstuck, and you understand why voice and whole-manuscript structure are a different kind of problem that most novelists still do not trust it with. You have a short, written policy that is actually yours.
Keep this workbook. It is yours to keep, print and write in. Come back to chapter 6's policy whenever a new tool tempts you to hand over more than you meant to.
https://trust-agent.ai/exam/writing-fiction-with-ai-what-actually-helpsThis is free educational content published by Mickai LTD, not professional, legal, financial or compliance advice, and Mickai accepts no liability for how it's used.
Trust Agent publishes workbooks like this one for free because Mickai believes a writer deciding what to hand to AI, on the actual evidence rather than hype in either direction, is exactly the kind of informed trust this project exists to build.
Keep learning
The complete workbook
This workbook gives fiction writers, hobbyist or working, an honest answer to where AI actually helps a novelist and where it does not. You will see the real, contested survey evidence on how many fiction writers use AI, learn the tasks it is genuinely good for, understand why most novelists still do not trust it with their voice or their book's structure, and leave with your own written policy for using it in your own practice.
- 01The honest evidence: what the numbers on AI and fiction writing actually sayRead here · 1 exercise
Before any technique, get the numbers straight, because they genuinely disagree, and understanding why they disagree is the first real lesson in this workbook.
- 02Where AI earns its keep: brainstorming and getting unstuckRead here · 1 exercise
The single most consistently reported use across the surveys in this workbook is not writing prose. It is helping a writer think through options faster than staring at a blank page alone.
- 03Research and getting the facts right, without losing your reader's trustRead here · 1 exercise
When novelists in the Cambridge survey said they used AI, the most common reason was not creative at all: sourcing facts and information for a scene.
- 04Where most novelists still do not trust it: voice, structure and the whole draftRead here · 1 exercise
The same Cambridge survey found overwhelming resistance to AI touching a whole novel, and only slightly less resistance to a short section. This chapter looks at why voice and structure are a different kind of problem from brainstorming or fact-checking.
- 05Editing tools and the disclosure questionRead here · 1 exercise
Beyond brainstorming and research sits a narrower, genuinely useful category: tools built specifically for manuscript craft, and a newer question that comes with using any of them if you plan to publish.
- 06Your honest AI-writing policy, and the final exerciseRead here · 1 exercise
Everything so far comes together into a short, personal policy you can actually hold yourself to, then this workbook's closing page.
Also inside: a 8-point checklist, a glossary of 10 terms and 12 questions and answers to test yourself. 6 hands-on exercises, each with a worked answer at the back where the workbook gives one.
The PDF is free. We ask for your name and email before downloading. Read the privacy notice at the download form. Free to download and use for personal learning, study groups and inside your own team. Please do not resell the workbooks or republish them as your own. Link people to trust-agent.ai instead.
Get your free PDF workbook
Enter your name and email, then download the complete workbook. This form also works without JavaScript.
Test yourself
Questions and answers
How many fiction writers actually use AI?
It depends which question you ask. A November 2025 survey found 61% of working writers generally use AI at least sometimes, but within that same study only 42% of the 291 fiction authors surveyed did, and only 11% of them used it for text they actually publish.
Why do different surveys give such different figures for AI use among writers?
Mostly because they ask different questions of different groups: experimenting once is different from regular use, which is different again from publishing AI-touched text, and a mixed sample of professional writers answers differently to a fiction-only sample.
What is AI genuinely good for in fiction writing?
Brainstorming and getting unstuck: generating a spread of options, character motivations or plot directions for you to choose from or reject, rather than writing the finished version for you. This is the task every survey in this workbook finds writers reaching for first.
What makes a brainstorming prompt actually useful, rather than generic?
Giving it your character's specific situation and constraint, then asking for a set number of distinct directions with a named consequence for each, rather than a vague open request that has nothing real to work from.
Is it safe to use whatever fact an AI tool gives me for my novel?
No. A language model can state a wrong fact with the same confident tone as a correct one. Verify anything you would be embarrassed to get wrong, from a source you actually trust, before it reaches the page.
What did the Cambridge survey find novelists actually use AI for, when they do use it?
Mostly non-creative tasks: around 20% use it to source facts or information, and only around 8% use it for editing text, a much narrower use than the headline 33% who use it somewhere in their process might suggest.
Why don't most novelists trust AI with their voice?
A language model tends to smooth toward a statistically likely, competent, generic middle. Voice depends on specific, idiosyncratic choices held consistently, which is close to the opposite of what that smoothing produces.
How negative are novelists about AI writing a full novel or short section?
Extremely, in the Cambridge survey: 97% extremely negative about AI writing a whole novel, 87% about a short section, easing only for editing, where 43% remained extremely negative.
Do writers who already use AI trust its output?
Often not fully, even their own. A broader August 2026 survey of working writers found 71% had used AI, yet 60% of all the writers surveyed said they distrusted AI-supported writing, including work they had produced themselves.
What can a manuscript-editing tool actually judge well?
The mechanical layer: repeated words across a whole book, timeline inconsistencies, overlong sentences and pacing flags. It cannot judge whether a plot twist lands emotionally or a character's choice feels true.
Do I need to disclose AI use if I self-publish?
Many platforms now ask you to declare AI-generated or AI-assisted content at submission, but the exact rules differ by platform and keep changing. Check the current rules for wherever you plan to publish before you submit.
What should my own honest policy for using AI in fiction writing actually cover?
What you will use it for (brainstorming, sourcing facts you then verify), what you will never hand to it (your voice, whole-manuscript judgement), and a reminder to revisit the policy as tools and evidence keep changing.
Write as you go
My notes
Use this scratchpad for your own notes, working or answers as you read. Notes are saved in this browser on this device. Downloading makes a local copy; it does not send your notes to Mickai.
Consolidate your learning
Review your work
Work through the exercises and questions, then compare your reasoning with the worked answers. A scored exam and verifiable certificate are not yet available for this course.
Review the questionsLearn the language
Key terms
- Generative AI
- Software that produces new text, images, audio or video from a prompt, rather than retrieving something that already exists.
- Large language model
- The kind of AI system behind most chatbots and writing tools, trained to predict likely next words from huge amounts of text, which is also why it can sound confident while being wrong.
- Prompt
- The instruction you give an AI tool. A prompt with specific stakes, characters and constraints produces far more useful brainstorming than a vague one.
- Brainstorming (with AI)
- Using AI to generate a spread of options for you to choose from or reject, rather than asking it to write a finished passage you keep as written.
- Voice
- The particular, consistent way a writer's narrator sounds: word choice, rhythm and the specific things only that narrator would notice, held across an entire book.
- Hallucination
- A confident-sounding but false statement produced by an AI system, which is why any AI-sourced fact needs independent verification before it goes into a manuscript.
6 of the workbook's 10 terms. The complete glossary is in the workbook.
Follow the evidence
Sources and checks
Facts last checked: .
These workbooks use AI assistance. See how the workbooks are made.
- New Report Examines Writers' Attitudes toward AIPublishers Weekly
- Half of UK novelists believe AI is likely to replace their work entirely, Cambridge research suggestsUniversity of Cambridge
- Writers use AI and distrust it, even when the words are their own, Medium report saysGamesBeat
Created by Mickarle Wagstaff-Irons - Micky Irons with the Mickai team. Published by Mickai LTD. Last updated 30 September 2026.
NextKeep going
Where to go next
Recommended for you
Use AI to study, research and write without losing your own thinking
Use an AI assistant as a study partner, not an answer machine: quiz yourself, verify every claim and citation, write in your own voice, and say how you used it.
Recommended for you
Why does AI make things up? Hallucinations explained
A plain-English guide to why AI chatbots invent facts in a completely confident tone, a practical checklist for spotting one, and what genuinely helps.
Recommended for you
Your first prompts: how to ask AI for what you actually want
Learn to ask AI for what you actually want: the six parts of a good prompt, how to improve a reply, and how to build a personal prompt library. Free tools only.