Worldbuilding: How Much Is Enough Before You Draft?

Worldbuilding is a real craft and a real trap. The question is not "how much worldbuilding is good" but "how much do I need before the draft can start, and how do I know when I have it." This post gives you a test.

The test: could a plot point depend on it?

Before you build anything, ask: could a scene in this book turn on this fact?

If the answer is yes, build it now, and build the limit before the capability. If the answer is no, or "maybe, eventually," write it down as one line in the open-questions list and move on. You will build it on the day a scene needs it, and you will build it better then, because you will know what the scene needs.

Some examples of the test applied:

Most of a four-hundred-page world folder fails this test.

What "enough" contains

For most novels, "enough to start" comes to a few thousand words, organised under four headings. This maps directly onto section 3 of the story bible template.

1. Rules with costs (the systems)

For every system a plot point can depend on (magic, technology, law, money, religion, biology), you need four lines:

• How it works, plainly. One paragraph a reader could understand.
• What it costs. Nothing free. If magic costs nothing, it is a plot hole waiting to happen.
• What it cannot do. The hard limits. This is the line that makes tension possible.
• Who controls it. Because that is a source of conflict.

Notice what is not on the list: history, taxonomy, the names of schools, the founding master. You can invent all of that on the page when a character mentions it, and the bible will record it.

The order matters. Writers build the capability first because it is fun, and then discover in chapter twenty that nothing can threaten a protagonist who can do that. Build the limit first. Ask "what can it not do?" before "what can it do?"

2. Places, three details each

For every place your first act visits: where it is relative to other places (with travel time), three sensory details, and who is there.

Three details, not thirty. A reader remembers the smell of the archive (dust, mouse, old wax) and the way sound dies in it. They do not remember the square footage or which century the east wing was added. If a scene needs the east wing, you will build the east wing that day.

3. One institution per conflict

Every conflict in the book pushes against something: a government, a guild, a church, a family, a company. For each one the protagonist will collide with in the first act, you need what it wants, who leads it, what it thinks of your protagonist, and how far its reach extends.

That is a paragraph each. It is not an org chart.

4. The daily texture

This is the one most systems-minded worldbuilders skip and most readers care about: what does a normal day look like for the protagonist? What does she eat, what does she owe, what is she afraid of getting caught doing? Texture is what makes a world feel inhabited rather than designed, and it comes from the character's routine, not from the world's history.

The two failure modes

Too much: worldbuilder's disease

The symptoms: you have a map but no scene list; you know the tax structure but not what the protagonist wants; every question about the story becomes a question about the world.

The underlying problem is that worldbuilding produces pages without producing risk. A world cannot be badly written. A scene can. So the anxious writer builds.

The cure is mechanical. Set a limit (two weeks, or three thousand words), write only the four categories above, and start the draft with the open-questions list beside you. Everything you did not build will be built on demand. Nothing you needed will be missing, because the scene will tell you what it needs.

Too little: the convenient exception

The opposite failure: no rules, so every problem is solved by inventing a capability. The healing spell that appears when someone is dying. The travel that takes an afternoon when the plot needs speed and four days when it needs delay. Readers of fantasy and science fiction are unforgiving about this, because the genre's pleasure is that the rules are real.

The cure is the same four lines: cost, limit, control, and writing down every exception the moment it appears in the draft, with a chapter number, so it becomes a rule.

A before-and-after on the page

Worldbuilding also fails on the page, in the form of the info-dump. Here is a draft opening from a workshop, details changed:

The Kingdom of Aldris had been founded four hundred years earlier by the Seven Houses, who had signed the Concord of Thorns after the Long Winter. Under the Concord, succession passed through the eldest child of the reigning monarch, authenticated by the Royal Archive, whose Keeper held the Seal of Record. The archive itself, a vast building of grey stone, had been expanded three times and now held over a million documents.

Everything here is true and none of it is a story. The reader has no reason to care about the Concord of Thorns, and the paragraph has told them the succession rule instead of making it matter.

The rewrite, applying the test (what does this scene depend on? the succession rule and the seal) and the daily-texture principle (what does Mara notice?):

The succession box had not been opened in Mara's lifetime. She knew where it was because she had dusted around it for fifteen years, and she knew what was in it because that was her job, and she knew the seal that would authenticate its contents was not in the archive, because she had checked, twice, on the night the heir died. The courier was still waiting at the door. She locked the box again.

The reader now knows: there is a succession document, it needs a seal, the seal is missing, the heir is dead, Mara is the kind of person who checks twice and locks things. The Concord of Thorns can wait. It may never come.

How much for each genre

The test is the same; the answer differs.

• Contemporary fiction: almost none before drafting. The world is ours. You need the specific institutions (the hospital, the firm, the school) and the protagonist's routine.
• Historical: the daily texture and the institutions, researched. The plot-dependent rules are real ones, so you cannot invent them; you look them up when the scene needs them, and keep the sources in the bible.
• Secondary-world fantasy: the four categories, with the magic system's limits written before its powers. Maps are optional; travel times are not.
• Romantasy: the rules that structure the romance (courts, bonds, prohibitions, the thing that makes them unable to be together) matter more than the rules of the war. See the romantasy guide.
• LitRPG and progression fantasy: the system is a character. Its rules need to be written and fixed early, because readers track them like a spreadsheet. See the LitRPG guide.
• Science fiction: the one or two technologies the plot depends on, with their limits. Everything else is set dressing.
• Web serials of any genre: the same, plus the discipline of recording every invented fact in a chapter summary, because you cannot go back and fix chapter 12 once it is published.

When to build more

Build more when a scene asks a question the bible cannot answer. That is the only trigger you need. It will happen constantly in the first act and rarely by the third, which is the sign that the world is now the right size for the book.

Some writers like to build between drafts instead: draft one with minimal world, note every question the draft raised, build the answers, revise. This works well for discovery writers and badly for anyone who cannot stand a draft with holes in it.

Worldbuilding with AI

Models are excellent at generating worldbuilding, which is exactly the problem: they will produce ten pages of provinces and holidays in a minute, and it will all fail the test. Use them for the opposite job. Paste your rules and ask: "What is the most convenient exploit of this magic system? What would a clever villain do with it? Where is the rule ambiguous?" That produces limits, and limits are what you need.

For places, ask for ten sensory details and keep three. For institutions, ask what the institution would do if the protagonist did X. You are using the model as a hostile reader of your world, not as its architect.

The checklist to stop and start drafting

You have enough when you can answer yes to all of these:

1. Every system a plot point could depend on has a cost and a hard limit written down.
2. Every place in the first act has a location, a travel time and three details.
3. Every institution the protagonist collides with in the first act has a want and a leader.
4. You can describe the protagonist's ordinary day in a paragraph.
5. Everything else is on an open-questions list, not in a document.

If you answered yes, close the world file and open the manuscript. The world will finish itself while you write.

Try it in SparkLore

SparkLore's bible has a world-rules entry type with the four fields (how it works, cost, limits, control) built in, which is a small thing that turns out to matter: it is hard to write the capability without being asked for the cost. If you are still testing the shape of a premise, the free AI plot generator will hand you a three-act skeleton, and the scenes in that skeleton are a fast way to find out which world facts a plot point actually depends on. Start a project with the open-questions list already open.