Worldbuilding without the wiki trap
· The StoryProp team
There is a writer — every fantasy reader knows one, and some of us have been one — who has spent years building a world and no time writing the book. The map is beautiful. The dynasties are numbered, the calendar has intercalary days, the language has a case system. Somewhere in the notes there is a document called “Chapter One,” and it is empty. This is the wiki trap: a body of lore that grows because growing it is pleasant, attached to a novel that never starts.
The trap is comfortable because worldbuilding is real work that carries no risk of being bad on the page. An encyclopedia entry cannot have clumsy rhythm; a trade route cannot be faulted for weak characterization. Lore accumulates the feeling of progress without ever exposing you to the two hard problems of fiction, which are scenes and sentences. And the tools flatter the habit: a personal wiki rewards completeness, cross-references, tidy categories. A novel rewards none of those things. A novel rewards a person walking into a room where something is wrong.
How to build a fictional world: start with the next scene
The way out is to invert the order of operations. Instead of building the world and then writing the story in it, write the story and build exactly what the next scene requires. If the scene is a smuggler bribing a customs officer at a river gate, you need to know the tariff on what she is carrying, when the gate changes watch, and who the officer is afraid of. You do not need the empire's full tax code, the history of the gate's construction, or the succession that installed the current governor. You need three facts, and the scene will tell you which three.
This can sound like a downgrade — worldbuilding for fantasy novels reduced to improvisation — but in practice it is a filter, not a shortcut. Every invented fact has to earn its place by doing work in a scene, which means the world you end up with is made entirely of load-bearing material. The wiki-first approach generates lore and hopes some of it becomes relevant. The scene-first approach generates only relevance, and the lore that accumulates behind it is, by construction, the lore of your actual book.
The iceberg is real — you only carve the tip
The standard objection is depth. Doesn't a world feel thin if the author hasn't worked out what lies under the surface? The iceberg image writers pass around — the visible tip supported by the submerged mass — is true, but it is widely misread. The submerged mass does not have to be written. It has to be implied, and implication is cheap if your details agree with one another. A border guard who spits at the mention of the old capital implies a war. If, two hundred pages later, a tavern song curses the same city, the reader now believes in a history you never drafted.
What actually needs to exist below the waterline is not text but commitment. A handful of decisions — the old kingdom fell, the river people were blamed, the blame was convenient for someone — will radiate through every scene they touch, and readers assemble the rest with a generosity no appendix can match. Depth is an experience the reader has, not a word count the author reaches.
Rules make worlds feel true. Inventories don't.
An inventory is a list of things that exist: twelve kingdoms, nine orders of mages, six ages of the world. Inventories are inert. A rule is a statement about what things cost, and rules generate story on contact. Say magic in your world is paid for in memory — small workings cost trivia, great ones cost faces and names. That single sentence produces a healer who no longer recognizes her daughter, a black market in strangers' childhoods, rich families who hire casters rather than let their heirs learn, and a reason your protagonist hesitates at exactly the wrong moment. Twelve kingdoms produce a map.
Rules do something else: they make the world checkable. Once the cost of magic is fixed, a reader could catch you cheating — and knowing you could be caught is precisely why they trust you. The generative questions in worldbuilding are rarely “what else exists?” They are “what does this cost?” and “who does it inconvenience?” A world where wonderful things have prices behaves like a real place. A world where the author can always reach for one more wonder behaves like a menu.
Reveal the world by consequence, not tour
The wiki trap has a prose style, and the style is the guided tour: paragraphs where the story stops so the world can be presented. The alternative is to let the world reach the reader through what it does to people. Suppose the pass at Coldgate closes at first snow. The touring version explains the geography of the mountains. The consequence version shows a messenger doing arithmetic in her head — days left, distance remaining — and lying to her captain about the answer. Characters, meanwhile, should almost never explain their world to each other. Nobody explains money to a cashier. The moment two locals discuss how their own magic works for the reader's benefit, the world stops being a place and becomes a brochure.
None of this is exclusive to fantasy. A realist novel builds its world the same way. Take a realist manuscript — call it “The Quiet House” — where Mara's town has an economy (the mill closed a generation ago), a geography (the highway that bypassed Main Street), and artifacts with history (a chipped cup that has outlived the marriage it came from). None of it was designed in advance; each fact arrived because a scene needed it, and each is revealed by consequence — who left town, which shops survive, why the cup is still in the cupboard. The method is genre-general. Fantasy just turns up the volume.
Once you've said it, it's canon
Building on demand carries one obligation the wiki approach handles by default: facts invented on the fly still bind. If chapter three says the pass closes at first snow, the relief army in chapter twenty cannot march through it in midwinter — not without turning your geography into a rumor. So keep a short ledger of what the text has actually stated: geography and distances, the costs and limits of magic, named history, who knows what. The useful distinction is stated versus planned. Stated facts are in the manuscript and are now load-bearing; planned facts are still yours to change. The broader craft of keeping a long story consistent is its own subject; for worldbuilding the point is narrower — the ledger records what you have committed to, and nothing else.
This is also the honest test of any AI worldbuilding tool. Almost anything can brainstorm a magic system; a general chatbot will invent one with you happily. The question that matters over a long project is what happens to the facts you settle on. An AI writing tool for fantasy novels should treat a settled fact differently from a floated idea. StoryProp draws that line explicitly: decisions you agree to are promoted to records — canon the agent reads before drafting anything — while everything you merely discussed stays a suggestion. Research works the same way, saved with sources and dates and kept apart from invented canon, so the real medieval salt trade never quietly bleeds into your imagined one.
A modest wiki is not the enemy; a ledger of stated facts, a map or two, a page of rules will serve any project. The trap is only the inversion — lore as the work, chapters as the someday. Chapters generate lore better than lore generates chapters, because scenes force the questions that matter: what does it cost, who does it inconvenience, what fell, who was blamed. Open the draft. Ask what the next scene needs to be true. Build that, write the scene, and let the iceberg grow under the story, where it belongs.
The scene-first method has one dependency: the facts you invent have to stay invented. StoryProp is built around that line — when you decide the pass closes at first snow, the decision becomes a record the agent reads before every draft, and what you didn't decide stays open.