Last reviewed
Worldbuilding is a craft most writing courses gloss over. There’s the “do you write a synopsis first or pants it?” debate, the “outline vs discovery draft” debate, the “POV character or omniscient narrator?” debate. But the practical question — what should I actually write down about my world before the manuscript starts, and how should I keep track of it once it’s started? — gets answered by the same hand-wavy “write everything down” advice that’s been useless to every writer who’s ever tried to follow it.
This pillar is the long-form version of the practical answer. It draws on how writers who actually ship work — Tolkien, Sanderson, Le Guin, Pratchett, Pierce, on the prose side; the lore teams at Riot, Bethesda, Bungie, FromSoftware on the game side — handle the part of the job that happens before and during drafting. Some of it is technique. Most of it is process. None of it is “buy a worldbuilding workbook.” The argument is structural: a world that holds together isn’t built by writing more notes; it’s built by writing prose against a structured reference that keeps you honest.
If you want the short version, start with What Tolkien, Sanderson, and game designers can teach you or Why your story bible isn’t enough. If you want the tooling story underneath, that’s PlotLens, and the conceptual frame is What is narrative intelligence?.
What worldbuilding actually has to do
Useful working definition: worldbuilding is the discipline of making the setting feel like it exists independently of the story you’re telling in it. It’s not the same as research, though research feeds it. It’s not the same as outlining, though outlining can lean on it. It’s not the same as a story bible — a story bible is what you produce by accident if you do worldbuilding right.
The job has three loads to carry:
The reader-immersion load. A reader believes the world when the details cohere. Marketplaces have currencies; currencies imply economies; economies imply where the food comes from, who makes the boots, how the roads get paid for, who patrols them. You don’t have to show all of it — most of it stays off the page — but the writer has to know enough that the on-page details point to something real underneath. The reader’s brain pattern-matches against the off-page implications constantly, and a writer who doesn’t have answers gets caught.
The author-decision load. Once a world is built, decisions inside the manuscript become local. “Can my protagonist hire a courier in this city?” becomes a one-second answer because you already decided how couriers work, how cities like this one operate, what hiring looks like. Without that structural decision, every micro-question stalls the draft.
The continuity load. This is the load that swamps everyone the first time. A world that’s only in your head will drift; a world that’s in your head plus a notes file will drift more slowly; a world that’s derived from the manuscript can’t drift, because the manuscript is the world. (We come back to this in the “Bible vs prose” section below.)
Most writers focus on the first load and ignore the third, which is why most worldbuilding projects fall apart somewhere around manuscript word 60,000. The third load isn’t separable from the first two; it’s the structural condition for the first two to keep working as the project scales.
How the masters actually do it
The popular stories about worldbuilding heroes are misleading. Tolkien didn’t sit down and “write Silmarillion first, then Lord of the Rings”; he spent forty years inventing Quenya and Sindarin because the philology was the personal joy, and the novels emerged from the linguistic infrastructure as a by-product. Sanderson doesn’t “write twenty pages of magic-system rules before drafting”; he writes the magic-system rules he needs for the scenes he’s about to write and revises them when scenes argue with them. Le Guin started The Left Hand of Darkness with a question (what would a society without fixed gender feel like?), built outward from that question for a decade across multiple books, and let the world fill itself in as the question pulled on its threads.
What’s actually consistent across these writers — and across the modern game-studio worldbuilders, who do this work on shorter cycles for larger teams — is a small set of habits:
Build to a question, not to a quota. “I’ll write 10 character bios and 5 city descriptions” produces inert reference docs. “What does the magic system have to be true for this scene to work?” produces structural decisions you actually use. Worldbuilding driven by writing questions makes the world directly responsive to the manuscript; worldbuilding driven by completionism produces archives the writer never opens.
Decide the constraints early; the consequences come later. Sanderson’s “second law of magic” — limitations are more interesting than powers — generalises. The most useful early-stage worldbuilding decisions are constraints: what can’t happen, what costs what, what’s off-limits, where the gravity is. The shape of the world follows from the shape of the constraints. Front-loading consequences (every magical creature, every guild, every faction) without first setting the rules produces a world full of detail and no internal logic.
Let the prose argue with the bible. A world is a hypothesis. The manuscript is the test. When a scene insists on a detail that contradicts the worldbuilding doc, the question to ask is “is the scene wrong or is the worldbuilding wrong?” Often the worldbuilding is wrong, because the scene is being honest about what the world has to be like for the character to act this way. Writers who let the prose update the bible end up with worlds that feel alive. Writers who insist the bible governs the prose end up with worlds that feel like reference manuals.
Stop at the threshold of usefulness. Tolkien’s languages have a complete morphology because he wanted them to. You don’t need yours to. The threshold to clear is “does the on-page detail feel grounded?” not “does the off-page lore reach completionist standards?” Worldbuilding to the wrong threshold is the most common form of procrastination in fiction; it produces the worldbuilding paradox where the most invested writers ship the slowest.
These habits aren’t compatible with the worldbuilding-as-form-filling workflow most worldbuilding software encourages. The premise of “fill in this template for a magic system, fill in this template for a faction, fill in this template for a calendar system” is form-completionism, not craft. The writers who do this well treat the bible as a working document — a place to record decisions as they get made — not a content artefact to populate before the writing starts.
What to track (and what not to)
A short list of what’s worth structuring, by category:
Characters — names, aliases, physical description, age, voice/dialogue patterns, what they want, what they know, what they fear, what they’re capable of. The “what they know” column is the one most systems miss; see Belief vs fact for why it matters.
Locations — name, type, what’s nearby, what’s distinctive, who lives there, what the on-page descriptions establish. Locations need spatial relationships (“the inn is on the river side of the market square”) because spatial drift is the kind of continuity error that’s hardest to spot in a re-read.
World rules — the constraints. How magic works, how technology works, what the laws of the land are, what the laws of the universe are. Rules are the load-bearing part of the worldbuilding; everything else is consequences. A change to a rule ripples through every scene that depends on it; a change to a city description ripples through only the scenes set in that city.
Factions and institutions — political parties, guilds, religious orders, criminal networks, corporations. For each: who’s in it, what they want, what they hate, what they pay/charge for, what their public face is, what their private operation is.
Timeline — explicit dates, durations, ages, season-of-year cues. The single most useful piece of structured data for a fantasy series is a timeline that knows every character’s age at every plot beat.
Languages, calendars, currencies — only if they’re load-bearing for the story. A culture’s calendar matters if a scene depends on a particular festival or season; otherwise it’s flavour. The threshold is “would a reader notice if this were wrong?” not “would a worldbuilding completionist be embarrassed by what’s missing?”
What’s not worth tracking:
- Anything that doesn’t show up on the page or constrain something that shows up on the page. If a fact has no on-page consequence, it’s not part of the world; it’s part of your imagination, and imagination doesn’t need filing.
- Anything you can derive from something else you’ve recorded. Don’t store the same fact twice; derive it.
- Anything that’s the same as the real world. Default to Earth; record only the divergences. This sounds obvious and almost no-one does it; new writers tend to specify gravity, sun colour, breathable atmosphere, hours in a day for fantasy worlds whose entire point is that none of those vary from Earth. Let the defaults default.
Bible vs prose
The single most important structural question in worldbuilding is: is the bible the source of truth, or is the manuscript?
Most worldbuilding tools assume the bible is. You write the bible first, the bible governs the prose, and contradictions are errors the writer introduced into the prose. This works fine for the first novel. By novel three of a series, the bible drifts from the prose every time a revision pass changes a sentence, and the writer is now maintaining two artefacts that disagree with each other in ways nobody catches until a reader does.
The structurally better answer is: the manuscript is the source of truth, and the bible is a generated view over it. Every fact in the bible points back to the sentence in the prose where it was established. When the prose changes, the bible changes; there’s nothing to maintain because the bible isn’t an artefact, it’s a query. This is the inversion PlotLens is built around, and it’s also the reason ad-hoc worldbuilding systems — spreadsheets, Notion, World Anvil — eventually fail at scale: they assume the wrong direction of authority.
For early-stage worldbuilding, the bible-first direction is fine. You’re inventing the world, the manuscript doesn’t exist yet, so the bible has to be the source. The moment the manuscript exists, the direction should flip. Notes you made before the first draft become hypotheses the draft is testing; notes you maintain after the first draft are duplicates of what the prose already says.
The shorter framing: worldbuild forward, validate backward. The bible’s job before drafting is to capture decisions. The bible’s job after drafting is to reflect what the prose actually says. The same artefact can serve both phases, but only if the second phase generates the artefact rather than maintaining it.
Tools and their limits
A pragmatic survey:
Notion / Obsidian / spreadsheets — flexible, infinitely shapeable, and the default for most writers. The wins are control and portability; you own the files, you decide the structure, and nothing’s stopping you. The losses are everything in the previous section: the bible drifts, the manuscript isn’t connected to it, the only enforcement is your discipline. For early-stage worldbuilding before there’s a manuscript to validate, this tier is fine and arguably best. Once the manuscript is going, the maintenance cost grows linearly with manuscript length and never stops.
World Anvil / Campfire / Plottr — purpose-built for fiction, with structured templates for the usual categories. The wins are guided structure and decent community/template support. The losses are the same: still bible-first, still requires manual maintenance, still drifts. Some have light cross-referencing; none of them read your manuscript and validate against it. Useful for the worldbuilding-forward phase, less useful for the validation-backward phase.
Scrivener — drafting environment with metadata. Holds metadata fields per scene, lets you tag characters and locations, supports a corkboard view. Strong for keeping the manuscript and the metadata in the same artefact, but the metadata is what you typed in, not what the prose says. Same drift problem at scale, just slower.
AI-drafting tools (Sudowrite, Novelcrafter) — bring generation to the worldbuilding party. Strong at “give me 10 names for a desert city” or “rewrite this magic system description in a different voice.” Less strong at validating that what you wrote yesterday agrees with what you wrote six months ago — that’s a retrieval and reasoning problem, not a generation problem. Useful for one phase, less useful for the other.
PlotLens — explicitly built for the validation-backward phase. Reads your manuscript, extracts every entity, builds the relationship graph, and validates new content against the established canon with sentence-level citations. Doesn’t help you invent the world; helps you ship the manuscript that was supposed to use it. Pairs naturally with any of the bible-first tools above for the early-stage phase. How PlotLens tracks every fact is the engineering walkthrough.
The honest position: most working writers benefit from one tool in each phase. A notes-style tool for the worldbuilding-forward decisions (Notion, Obsidian, or just a single text file is fine), a generative tool if generative AI helps you draft (Sudowrite, Novelcrafter), and a manuscript-derived continuity layer for the validation-backward phase (PlotLens). The mistake is using one tool for all three jobs and asking it to do work it wasn’t built for.
Stop building, start writing
Worldbuilding is fun. Drafting is hard. Almost every writer who loves the worldbuilding hits a moment where they realise they’ve spent more time on the wiki than on the manuscript, and the wiki is what they’re producing instead of the book. There’s a clinical term for this — worldbuilding rabbithole syndrome — and the only honest treatment is “stop building, start writing.”
The signals that you’ve crossed the threshold from useful worldbuilding into procrastination:
- You’re worldbuilding categories that don’t constrain a scene you’re about to write. (Inventing 14 species of tree in a forest the protagonist crosses on horseback over the course of one sentence.)
- You’re building the bible to a completionist standard rather than a usability standard. (“Every faction needs a 2,000-word lore page” is a completionist standard; “every faction needs enough rules that I know how members of it act in a scene” is a usability standard.)
- You haven’t shipped a manuscript word in two weeks.
- You’re more excited to open the worldbuilding tool than the draft.
The threshold is reached. Open the draft. The rule almost no-one follows: worldbuild what you need for the next scene, not the whole world. The rest of the world emerges as the prose demands it. The rest of the worldbuilding can wait until the rest of the prose exists.
A useful prompt when stuck: “What does my world have to be like for this scene to work?” Worldbuild the answer. Write the scene. Move on. The bible accumulates by necessity, the manuscript accumulates by progress, and the gap between writers who ship and writers who don’t is mostly the willingness to keep the worldbuilding subordinate to the prose.
Where to go from here
If you’re at the start of a project, How to build a story bible is the practical companion to this pillar — the same argument turned into a step-by-step. If you’re deep into a manuscript and the worldbuilding is starting to drift, Why your story bible isn’t enough is the diagnostic for the structural problem. If you’re working on a series specifically, PlotLens isn’t another AI writing tool — it’s your story’s memory reframes the worldbuilding question as a memory question, which is the framing that actually scales.
The shorter version of this entire pillar: build forward to capture decisions, validate backward to keep the prose honest. The bible isn’t the world; the world is in the manuscript. Worldbuild what the next scene demands, stop where the prose stops needing it, and let the tooling derive the bible from your own sentences rather than asking you to maintain a parallel artefact. Worlds that hold together are built that way.