How to make a D&D campaign wiki your table will actually use

A campaign wiki is a record of play, not a setting bible

Two documents get called a campaign wiki, and they want opposite things. A setting bible is the world before play: the pantheon, the map, the thirty NPCs you wrote in July. A campaign wiki is the record of what happened once the players arrived: which of those NPCs the party met, what the table called them, what was promised, and which of your plans burned down in session four. The first mostly stays still. The second does not exist until the dice start, and it changes every week.

They get confused because both are "the campaign", and because the tools most DMs reach for were built for the first job. World Anvil and LegendKeeper are worldbuilding platforms: you author a world in them, and they are good at it. Nothing about them is wrong. It is just that writing the world is usually the thing you had already done before you opened them.

So the typical wiki is born as a setting bible and then quietly asked to take on the second job once play starts. The lore pages stay pristine while the one page that matters — what the party did last week — never gets written, and the DM concludes they are not a wiki person. They were maintaining the wrong document. If your setting bible already lives somewhere you like, leave it there; the two do not have to share a home.

The six pages every D&D campaign wiki needs

Strip a campaign wiki down to what gets opened mid-session and six kinds of page survive. Everything else — house rules, the pantheon, the calendar — is reference, and reference does not go stale. These six do, because every session changes them.

• NPCs: one page per person the party has actually met — the name the party uses for them, what they want, what they know, what was promised in either direction. The NPC tracker template is a copyable card for exactly this.
• Places: one page per location the party has been, not per location on the map. What it looks like from the party's side, who they met there, what they left unfinished. A town they have never visited is setting bible, not wiki.
• Factions: who they are, where they stand toward the party right now, and what changed that stance, with the session number. Attitude drifts most between sessions and is noticed least until it matters; the faction tracker template is a sheet built around that one line.
• Items: the ones that carry plot or power — the letter, the key, the sword with a name. Who holds it now, where it came from, what it has been used for. Mundane loot stays on the character sheets.
• A session log: one entry per session, numbered and dated, written once and never edited. It is the campaign's history, and the campaign notes sheet has the ten-minute post-session template for it. (On paper, D&D Beyond publishes a free Campaign Journal among its tracking sheets — sourced in the table at the end.)
• An open-threads page: every unresolved hook, promise and question, with who cares about it and when it was last touched. One page you rewrite rather than add to — the open-threads list in the campaign memory guide shows how to keep it from becoming a graveyard.

How to structure an entry so it survives session twenty

An entry that is easy to write in session two and useless by session twenty usually fails in one of five ways. First, one page per thing. A page called "The Docks" holding the harbormaster, the smugglers and the warehouse fire matches how the scene played — and by session twenty the harbormaster has turned up in three other places and her facts are scattered across three pages. Give her a page, give the docks one, and link them.

Second, put the party's name for a thing on the same page as its real name. Nobody says Sera Vell by session nine; they say the nervous one, and if that lives only in your head you will search for the real name, forget it, and create a second entry — the usual way a wiki gets two pages for one person. Third, one spelling per name: Thordak or Thordack, decided the first time you write it, because a search for one will not find the other.

Fourth, stamp every entry with first-seen and last-seen session numbers. A session number is the index into everything else — the log, your notes, the recording if there is one — and an entry without one is a fact with no date. Sort by last-seen before you prep, and the bottom of the list is the NPCs your players have forgotten.

Fifth, keep a block on every entry that players never see, and keep it off anything they can open. What the NPC secretly wants, what the faction is really planning, what the item actually does: the lines most useful to you and most damaging on a shared screen. Decide once where that line sits, and never write a secret above it.

Four ways to build one: worldbuilding platform, campaign manager, notes app, or generated from recordings

There are four honest routes, and they differ less on features than on who does the typing. Every price lives on the relevant comparison page, dated and linked; none is repeated here, because a price in a sentence is wrong the week after it changes.

A worldbuilding platform — World Anvil or LegendKeeper — is the right pick if you are an author first: you enjoy writing the setting, and the record of play will be one wing of a larger building. It is the most structured option, and the one that most needs the ten-minute habit below.

A campaign manager such as Kanka is shaped like a campaign rather than a world — characters, locations, organisations, journals and the relations between them — and it has a genuine free tier, sourced in the table below. If you want to track play rather than author a setting, it is the closest fit among the hand-built options.

A notes app is the route most DMs are already on, and Obsidian or Notion cover the two shapes: Obsidian keeps the wiki as plain files on your own disk, with good linking and backlinks; Notion is for a group that edits together, with database views that make an NPC table pleasant (its free plan caps a file upload well below any session recording — the figure is in the table). Both give the least structure and the most freedom, which suits a DM who already knows what they want.

The fourth route is to generate the wiki from the sessions themselves: record the game, run the recording through a tool built for tabletop, and the NPC, place, faction and item entries are extracted from what was actually said. DM Scribe, which publishes this guide, is one of these; the comparison hub lists the direct competitors. It is the only route where the wiki keeps up without the DM typing — and the only one that is useless if your table does not record.

The ten-minute habit that keeps a wiki true

Whichever route you pick, a wiki is only as true as its last update, and the update has a window. Within an hour of the session you still remember the innkeeper's name and the price you quoted for passage; by Wednesday you do not. So: ten minutes, straight after the session, before you drive home — a brain-dump from your in-session notes into the pages the session touched, not a polish pass.

Touch only those pages. A wiki dies when updating it means auditing it — walking every entry to see what might have changed. Work from the notes instead: every new name is an NPC page to create or stamp, every promise is a line on someone's entry, every open question goes on the open-threads page. If the session did not mention the Thieves' Guild, its page stays closed. A session only changes a handful of pages, which is why ten minutes is enough.

Delete rather than strike through. A resolved thread comes off the open-threads page entirely; a faction that changed its mind gets its line rewritten, not appended with "but now". History lives in the session log, the one page you never edit; every other page describes now. A wiki full of crossed-out lines cannot be read at a glance, and at a glance is the only way anyone reads it mid-session.

And give the session-log page to a player. Rotate the job every session, cap it at five bullets per prompt, and have it finished before anyone leaves — the player scribe sheet is built for it. A player's notes record what the party saw, which yours cannot, and the gap between the two is where your next session lives.

What players should see, and what stays in the GM block

A wiki the players can read is worth far more than one they cannot: a player who missed three sessions catches up without a recap, and most of the "wait, who was that again?" questions come off your plate. But once one wiki holds both the party's view and yours, the split has to be structural — something the tool enforces, not something you remember.

The split is simple to state. Players see who they have met, where they have been, what they were told, what they promised, and what the log says happened. The GM block holds what an NPC actually wants, what a faction is really doing, which rumour was true, and the plan for the thread they have not found yet. If a line would change how a player plays, it goes below the line.

In a hand-built tool this is a permissions question — a private section, a page players cannot open, a second vault — and every tool on the routes above has some version of it. Check how yours draws the line before you write the first secret, because moving secrets later means rereading every page. In a generated wiki the split is built in: ours keeps a GM-only notes block on every entry that is never shown to a player, and read-only links let the table read the rest without editing it — on the Dungeon Master plan and above.

If you record your sessions, the wiki can write itself — and what it cannot write

Everything above is work the sessions already contain: every NPC name was said out loud, every promise was made at the table, every shift in a faction's stance was played. A recording holds all of it. On ours, each session is transcribed and the NPCs, places, factions and items it mentioned are extracted into entries in the campaign wiki feature, each with an Appearances list — which session it surfaced in, with the model's quote of how it came up — so first-seen and last-seen stop being lines you have to remember to write.

Be clear about what it cannot write. A transcript is text with no speaker labels, so an entry's quote tells you what was said, not who said it. Extraction is sometimes wrong — a table joke becomes an NPC, a sound-alike becomes a second entry — so low-confidence and possible-duplicate entries land in a GM review queue, and a per-campaign ignore list holds the names the AI must never turn into an entry again. Entries drafted from a recap or created by hand carry no Appearances list, because nothing was extracted for them. And the GM block stays yours: what an NPC secretly wants was never said aloud, so no recording holds it.

The plan facts, since they decide whether this route is open to you: the free Adventurer plan gives you 2.5 hours of transcription a month plus 3 one-time bonus hours at signup, for one campaign, with no credit card to start. Transcript import is on every plan — two a month free, ten on Dungeon Master, unlimited on Guild Master — so a transcript made elsewhere can build the wiki without spending hours. And if your table does not record, none of this applies; the five sections above are the whole system, and they cost nothing.