Table of Contents
Whodunits break writers because the genre demands two contradictory things at once. Hide the killer well enough that readers feel surprised. Plant every clue fairly enough that they feel cheated when they miss it. Miss that balance and your mystery either reads as obvious by chapter four or arrives at a solution that feels pulled from thin air.
The good news. AI handles the part of mystery writing that wrecks most drafters: tracking who knew what, when. Sudowrite's Story Bible was practically designed for this puzzle work. Its prose models can hold a fair-play structure across 80,000 words without losing the gardener's alibi.
Why Whodunits Are the Hardest Genre to Draft Alone
A romance can survive a soft middle. A fantasy can absorb a contradictory worldbuilding detail. A whodunit cannot survive a continuity error.
If your detective spots blood on the carpet in chapter six and the killer was supposedly out of town until chapter eight, your reader is done. They've spotted the seam. The spell is broken.
Mystery writers keep three things on their desk: a timeline, a knowledge matrix, and a clue tracker. Forget to update one and the manuscript drifts from its logic.
This is exactly the bookkeeping AI excels at. Not the imaginative leap. Not the moment your sleuth realizes the typewriter has a sticking Q key. The boring continuity work that makes those moments land.
The Fair-Play Rule You Actually Have to Follow
S.S. Van Dine wrote his twenty rules in 1928. Ronald Knox added his ten commandments a year later. Modern writers ignore most of the prohibitions, but the core principle survives: the reader must have access to every clue the detective uses to solve the case.
No secret twin reveals. No poison only the detective knows about. No off-page confessions. If your sleuth deduces from muddy boots that the killer is left-handed, those boots have to appear in scene with the print angles described in enough detail that an attentive reader could draw the same conclusion.
Agatha Christie cheated occasionally. Contemporary readers won't let you.
Setting Up Story Bible for Clue Tracking
Open a new Sudowrite project and head straight to Story Bible. Build out four sections before you write a word of prose: Characters, Worldbuilding, Outline, and Braindump. The Braindump field is where most of your mystery infrastructure will live.
Start with your suspect roster. Every viable suspect gets a Character card. Not just the killer. Every person the detective will reasonably consider. For each, set up these fields:
- Public motive. What everyone in the novel thinks they might have wanted.
- Hidden motive. What they actually wanted, which may or may not be murder.
- Alibi. Where they claim to have been at the time of the crime.
- Real location. Where they actually were, in tight prose.
- Secret. The thing they're hiding that has nothing to do with the murder. This is your red herring fuel.
- Knowledge state. What they know about the victim, the crime scene, and each other.
That last field matters more than people realize. What each character knows changes how they behave and what they accidentally reveal. A widow who knows her husband was poisoned reacts differently to wine than the widow who thinks he had a heart attack.
The Clue Manifest
In Braindump, create a numbered list of every clue you intend to plant. For each clue, write three lines: what it is, which chapter it appears in, and which characters notice it versus which miss it. Number them so you can reference them later.
A typical 80,000-word whodunit has fifteen to thirty plantable clues. About a third are red herrings pointing at the wrong suspect. Another third are ambient details that mean nothing until the reveal recontextualizes them. The last third are breadcrumbs leading to the killer.
This list needs to exist in Story Bible before you draft because every chapter gets checked against it. Sudowrite's models can hold this context when you point them at it. They cannot invent it for you.
Picking the Right Prose Mode for Mystery
Per Sudowrite's prose-modes matrix, Claude 3.7 Sonnet is the model for mystery work. Sonnet handles dialogue with restraint, doesn't over-explain inner thought, and writes atmospheric description without veering purple. The muscles a whodunit needs.
Muse, Sudowrite's fiction-trained default, is better for romance, erotica, and horror. It runs hotter on emotional intensity and won't refuse dark material. For cozy mysteries or procedurals, stick with Sonnet's cooler register. For noir with explicit content or psychological horror layered in, switch to Muse for the darker chapters.
Set your Creativity Dial to 4 or 5 for investigation chapters. You want prose that surprises you without inventing details that contradict your clue manifest. Crank it to 7 or 8 only for dream sequences, chase scenes, or paranoid interior monologue.
Tone Shift for Mystery Pacing
Tone Shift becomes your pacing instrument. Openings usually want Authoritative or Ominous. The middle, where suspicion mounts, calls for Conflicted. The final third often benefits from Fast-Paced as the detective assembles pieces. The drawing room reveal lands best on Authoritative with a touch of Ominous.
Mark these shifts in your Outline so you can apply them consistently when you Rewrite scenes that feel off.
Planting Clues Without Telegraphing Them
The first rule of clue placement: the clue cannot be the most interesting thing in the scene. Place it next to something more attention-grabbing and the reader's eye skips past it. Raymond Chandler called this misdirection. Wave the bright object in the left hand. The trick is in the right.
Here's a before-and-after. The detective interviews the housekeeper. Your buried clue: she polished the silver on Tuesday, but the silver tea service was missing from the police inventory on Wednesday. That contradiction is your clue.
The bad version places the silver detail front and center.
"I polished the silver Tuesday morning," she said. "All of it. The tea service especially. It hadn't been used in months."
Inspector Holloway made a note. The silver. He would need to check on that.
That's a flashing neon sign. Now compare:
"Tuesday's always polishing day," she said, twisting her apron. "Mrs. Ashworth was particular. I came in early because she'd asked me to stay late Wednesday for the dinner she never had. I told the constable about Wednesday. He didn't ask about Tuesday."
"And what did you polish?"
"The silver, the brass in the front hall, the candlesticks in the dining room. Took me till noon."
Holloway thanked her and moved on. The brass had been on his mind since he'd noticed the tarnished knocker on the front door. Someone had touched it recently, hard enough to leave a fingerprint in the polish.
The silver is mentioned in passing. The brass and the fingerprint pull focus. A careful reader remembers the silver later. A casual reader does not. Both outcomes are fine.
Using Write Guided for Clue Insertion
When a chapter needs to include three specific clues from your manifest, use Write in Guided mode rather than Auto. Auto follows the story's natural momentum. Guided lets you steer with explicit direction.
A useful Guided prompt: "Continue the gardener interview. He mentions repotting hellebores Wednesday afternoon. He notices muddy boot prints leading to the greenhouse but doesn't comment. End with the detective receiving a phone call."
You're naming the clue, the misdirection, and the scene exit. Sudowrite will draft the scene and surprise you with details you didn't ask for. Keep what works. Rewrite what doesn't.
Red Herrings That Actually Work
A red herring fails when the reader can tell structurally that it's a red herring. The common tell is the suspect who looks too obvious too early. Christie did this with butlers. Modern readers have absorbed the convention and now suspect the obvious suspect of being a setup, so your obvious suspect needs to be doubly false.
The most effective red herrings are characters hiding something real but unrelated. The doctor having an affair with the victim's daughter. The business partner embezzling small amounts. The neighbor concealing a previous arrest.
These secrets generate suspicious behavior and explain odd alibis. But they are not the murder.
When the detective uncovers the secret, the reader feels satisfied that something has been revealed. A chapter or two later they realize this person could not have killed the victim. That's the misdirection working.
Chapter Continuity for Red Herring Placement
This is where Chapter Continuity earns its keep. Run it after every two or three chapters. It flags contradictions you missed.
Say the gardener was repotting hellebores Wednesday afternoon in chapter four. In chapter nine, you have him meeting the victim's brother at the train station that same afternoon. Chapter Continuity catches that.
For red herrings, use Continuity to verify that your false suspect's secret stays consistent across every scene. The reader should be able to look back and see the suspicious behavior was always about the affair, not the murder. If you've left a thread dangling, Continuity will surface it.
Chat as Your Fair-Play Sanity Check
Sudowrite's Chat reads your Story Bible. That makes it the closest thing to a mystery editor you can summon at 2 AM. Use it as your fair-play auditor.
Questions to ask Chat after each major revision:
- Can a reader solve this murder from the clues planted before the reveal? Walk through the deductive chain. If Chat can solve it without referencing your reveal scene, your structure is sound.
- Which red herrings are weakest? Ask which false leads feel underdeveloped or which characters lack a believable secret.
- Where is the killer underused? A common whodunit flaw is the killer disappearing for the middle hundred pages. Chat can identify scenes where they should reasonably appear but don't.
- Do any clues only make sense in retrospect? A clue that requires the solution to be interpretable isn't a fair-play clue. It's an Easter egg. Both are fine in moderation, but you need at least seven or eight clues that work prospectively.
Chat won't catch every problem. It reads what you wrote, not what your reader will infer. But it catches enough structural weaknesses to fix before draft three.
The Whodunit Structure Checklist
Drop this checklist into Story Bible's Outline section as a template:
- The Body. Discovered within the first 10%. Specific enough to picture, vague enough that the killer isn't immediately suggested.
- The Sleuth. Introduced with a small case or quirk before the main mystery.
- The Suspects. Five to seven viable candidates, each with public motive, hidden secret, and one suspicious moment.
- The First Red Herring. Planted by the end of act one. Plausibility scales up through the middle.
- The Second Body. Optional. Eliminates one suspect around the 60% mark.
- The Detective's Wrong Turn. Mid-act-three. They pursue the wrong suspect with confidence.
- The Recontextualization. An earlier detail returns with new meaning.
- The Confrontation. The killer is named where they can react. Drawing room, study, or modern equivalent.
- The Explanation. Kept short. Lingering here drains tension.
Classic Fair-Play Failures and How to Fix Them
Three failures show up in unpublished whodunit drafts more than any others.
Failure 1: The clue is mentioned but never witnessed. A character "remembers" seeing something off-page. The reader has no way to verify it. Fix this by writing the original observation as a scene the reader experiences, even if the witness doesn't realize its significance at the time.
Failure 2: The motive only emerges in the reveal. The killer wanted the inheritance. Fine. But the inheritance was never discussed until the detective explains it.
Fix this by planting the financial angle early. The will, the lawyer's visit, someone complaining about money. The reader should know the motive exists before they know who acted on it.
Failure 3: The detective notices something the reader didn't see. Inspector Holloway picks up on the killer's wristwatch matching an impression in the dust. Lovely. But the wristwatch was never mentioned and neither was the dust.
Fix this by working backwards. Write the reveal first, list every detail the detective uses, then check that each one appears in earlier scenes witnessed by the reader.
Drafting the Reveal
The reveal is the hardest scene in a whodunit. You're delivering exposition, confrontation, and emotional payoff at once. Most writers over-explain.
Use Rewrite in Shorter mode on your first draft of the reveal. Almost always. Then ask which deductive steps the reader has already made. Cut those. Leave only the steps the reader couldn't have taken without the detective's specific knowledge.
For the killer's reaction, switch to Muse if your mystery runs dark. Muse writes the panic, the bargaining, the unraveling without flinching. Sonnet sometimes pulls its punches. Muse won't.
One technique from Tana French and Kate Atkinson: end the reveal scene before the killer is officially in custody. Cut to the next morning. Let the reader sit with the aftermath. The arrest is administration. The reveal is the story.
Putting It All Together
A working whodunit draft in Sudowrite looks like this. Story Bible loaded with five to seven Character cards, each carrying public and hidden motives. A numbered clue manifest in Braindump. A scene-by-scene Outline marking which clues land where.
Sonnet selected as your prose mode with the Creativity Dial at 4 or 5. Chapter Continuity run after every three chapters. Chat consulted as a fair-play auditor before each major revision.
You still write the prose. You still feel the moment in the gardener's interview when the misdirection clicks. But the machinery underneath, the tracking that breaks most first drafts, holds.
Whodunits reward writers who do the puzzle work upfront. The genre is unforgiving of shortcuts and generous to careful planners. Sudowrite has a free trial that lets you load a Story Bible and run a few chapters through the system. That's usually enough to see whether your mystery has the bones to carry eighty thousand words.