Random Prompt Generator for Writers: Tables and AI
September 14, 2026
A random prompt generator gives you a story idea you didn't pick: a character, a place, a problem, and ideally a twist. Most of them work one of two ways. Either they pull one entry from each of several lists and glue them together, or they ask a language model to invent something. You don't need a website for either. Below is a dice table you can roll right now, an AI prompt for ChatGPT or Claude, and a process for finishing the story.
The short answer: tables surprise you, AI writes smoother prompts, and the best setup combines them. Roll the ingredients yourself, then let the AI shape them.
How a random prompt generator works
The classic writers prompt generator is a set of lists: characters, settings, conflicts. The site picks a random line from each and prints a sentence. With five lists of 20 entries each, that's 3.2 million combinations from only 100 lines of text.
The idea predates the web. Brian Eno and Peter Schmidt's Oblique Strategies, first published in 1975, is a card deck you draw from when stuck.
AI generators have no list. The model writes a new prompt each time, which is flexible but changes what "random" means.
| List-based generator | AI generator | Dice table plus AI | |
|---|---|---|---|
| Where randomness comes from | A random pick from fixed lists | The model's sampling | Your dice |
| Surprise | High, sometimes absurd | Medium, drifts toward common ideas | High |
| Takes constraints (genre, tone, length) | Only if built in | Yes | Yes |
| Repeats itself | Only when lists are small | Often, across a session | Rarely |
Ask a chatbot for "a random story idea" ten times and the same lighthouse keepers keep coming back. The model isn't rolling dice. It predicts a likely answer, and likely answers cluster. Randomness from outside the model fixes that.
Build your own random prompt generator with dice
Roll one six-sided die for each column. Five rolls give you a prompt. The table is yours to use and edit.
| Roll | Character | Setting | Wants | Obstacle | Twist |
|---|---|---|---|---|---|
| 1 | A retired court interpreter | A ferry that never docks | To be forgiven by one person | A promise made to someone now dead | The narrator caused the problem |
| 2 | A night-shift pharmacist | A town about to be flooded for a reservoir | To leave without anyone noticing | A child who won't leave them alone | The antagonist is right |
| 3 | Twin sisters who haven't spoken in ten years | A wedding reception in a failing hotel | To prove a rumor false | A secret that would hurt someone innocent | The story is told backward |
| 4 | A debt collector with a conscience | A research station in the middle of winter | To finish something before a deadline | Their own body is failing | The goal is achieved and it's worthless |
| 5 | A teenage beekeeper | A border checkpoint | To learn what happened one night | The only witness is an unreliable friend | Two characters are one person at different ages |
| 6 | A widower learning to cook | A building emptied for demolition | To keep a family business alive | The law is on the other side | Nobody in the story is lying |
Example roll: 2, 5, 3, 1, 4. A night-shift pharmacist at a border checkpoint wants to prove a rumor false, is bound by a promise to someone who's dead, and succeeds, only to find the truth didn't matter.
To make the table last:
- Go to 36 rows. Roll two dice, read one as tens and one as units (11 to 66). Tabletop players call it a d66. Each column now holds 36 entries.
- Swap a column for your genre. Fantasy writers can replace Setting with "Magic that costs something." Romance writers can replace Obstacle with "Why they can't be together yet."
- Reroll one column only. If a result feels flat, keep five and reroll the weakest.
An AI writer prompt generator you can paste into ChatGPT or Claude
A good writer prompt generator prompt gives the model real randomness (your dice), hard constraints, and an example output. Anthropic's prompting best practices calls examples one of the most reliable ways to steer format and tone, and advises keeping them diverse. OpenAI's prompt engineering guide describes the same few-shot approach.
Paste this and fill in the brackets:
You are a writing prompt generator for fiction writers.
Ingredients (rolled on dice, use every one):
- Character: [from your roll]
- Setting: [from your roll]
- Wants: [from your roll]
- Obstacle: [from your roll]
- Twist: [from your roll]
Constraints:
- Genre: [literary / horror / sci-fi / romance / mystery]
- Length: [flash / short story / chapter]
- Point of view: [first / close third / second]
- Avoid: lighthouses, amnesia, chosen ones, "it was all a dream"
Give me 3 different prompts built from these ingredients.
Each prompt: 2-3 sentences, a working title, and the first line of the story.
Make the three prompts differ in tone: one quiet, one tense, one funny.
Here is an example of the quality I want:
Title: The Last Pickup
Prompt: A debt collector must repossess a piano from a widow.
He has one day, and he's already decided to lie to his boss.
First line: "Nobody warns you that pianos have to be tuned after a funeral."
The "Avoid" line blocks the ideas models reach for first. Add to it whenever a cliché shows up. For more on why structure like this helps, see what a prompt actually is and the patterns in our best ChatGPT prompts guide.
Writing prompt generator for adults
A writing prompt generator for adults should skip classroom prompts for grief, marriage, money, aging parents, and moral compromise, with endings that don't resolve neatly.
To get that, change the constraints, not the tool:
- Set Genre to "adult literary fiction."
- Add: "Themes: loss, obligation, compromise, divided loyalty. No moral at the end."
- Add: "Characters are over 35 unless the ingredients say otherwise."
Mature themes don't mean explicit content. The strongest adult prompts are uncomfortable because of the choices they force. Mainstream AI tools also apply content policies, so ask for depth, not shock.
Fanfic prompts generator: tropes, OCs, and fandom norms
Fan fiction has its own vocabulary. A fanfic prompts generator should use trope columns instead:
| Roll | AU (alternate universe) | Trope | Your original character's role |
|---|---|---|---|
| 1 | Coffee shop | Fix-it (repair a canon ending) | The one who knows a secret |
| 2 | Everyone's a medieval guild member | Hurt/comfort | A rival who becomes an ally |
| 3 | Modern-day office | Fluff | A new neighbor |
| 4 | Soulmates can hear each other's thoughts | Crossover | A witness to a canon event |
| 5 | Roles reversed (hero and villain swap) | Time loop | The canon character's estranged sibling |
| 6 | Canon, ten years later | Missing scene | Someone the plot forgot |
A few norms keep fan writing respectful:
- Center an original character (OC) when you can. It gives you room to invent without rewriting a canon character.
- Know the creator's stance. Wikipedia notes Anne Rice long opposed fan works based on her characters before softening.
- Tag and warn. Use ratings, content warnings, and a disclaimer crediting the original. Archive of Our Own is run by the nonprofit Organization for Transformative Works, which describes it as a non-commercial archive for transformative fanworks. Read the rules wherever you post.
- Leave other fans' stories out of your AI prompt. Use your rolls and canon facts, not another writer's fic, unless they agree.
How to turn a random prompt into a finished piece
Most random prompts die in a notes app. What turns one into a story is a small scope and a deadline.
- Decide the length before you start. Flash fiction under 1,000 words is the right size for a random prompt.
- Write the last line first. The twist column tells you where you're going.
- Draft fast, without AI. Use the model to generate prompts, not the prose.
- Use AI as a reader afterward. Paste your draft and ask: "Where did you stop caring? Which line would you cut?"
- Keep a log. Note which rolls produced stories you finished. Cut the columns that never work.
If you run this loop every week, save it instead of retyping it. A saved base prompt is one way to do that in a single chatbot, and a shared prompt library is the team version.
A prompt loop you've tuned over months is exactly the kind of thing that gets lost between chats. Taku lets you mirror a writing workflow someone already built into your own desktop workspace, remix it for your genre, and run it again next week. The marketplace's free apps are one place to start looking. Taku is in Beta, and the Mac app is available now.
Key Points
- Random prompt generators either combine picks from lists or ask an AI model to invent one.
- A dice table with five columns (character, setting, wants, obstacle, twist) gives you thousands of prompts for free.
- Feed an AI your dice results, constraints, an "avoid" list, and one example.
- "For adults" means mature literary themes, not explicit content. Change the constraints, not the tool.
- For fanfic, build tables from AU and trope columns, favor original characters, and follow your archive's rules.
- Finish by picking a short length, writing the ending first, and drafting without AI.
FAQ
Is there a free random prompt generator for writers?
Yes. The dice table in this post costs nothing, and neither does a spreadsheet with a random-pick formula. The AI prompt works in any chatbot you already use.
What's the difference between a writers prompt generator and an AI prompt generator?
A writers prompt generator produces story ideas. An AI prompt generator usually writes instructions for an AI model, like image prompts.
How do I get a writing prompt generator for adults without explicit content?
Ask for adult literary fiction and name the themes: loss, obligation, compromise, divided loyalty. Specify older characters and unresolved endings. That works better than asking for "dark" ideas.
Can a fanfic prompts generator use real characters from a show or book?
Fan fiction builds on existing characters by definition, so yes. Before publishing, check the creator's stance and the archive's rules, and add a disclaimer.
Why does ChatGPT keep giving me the same story ideas?
Language models predict likely text, and likely story ideas cluster. Give it outside randomness, like dice rolls, add an "avoid" list of clichés, and ask for several prompts that differ in tone.