Then session thirty-one happens. Your rogue swears the party still owes the smuggler a favor. Your cleric swears the smuggler owes them, because of the thing with the barge. The argument is entering its fourth minute. You open the wiki, the one you spent a weekend on, and it has a lovingly formatted page about the founding of the kingdom and absolutely nothing about the barge.

If you searched “dnd campaign wiki” wondering whether you need one and what goes in it, here is my answer up front: you probably do, it should be much smaller than you think, and almost everything the standard advice tells you to put in it is filler. A campaign wiki has exactly one job, and once you know what the job is, the whole thing gets easier to build and dramatically more useful to own.

The job: settle it in ten seconds

A campaign wiki is a dispute-settling machine. Its entire purpose is to answer a question, at the table, while a scene is running, in under ten seconds. That is the test. Who was the fence in Coppergate? Did the duke ever pay the party? What did the prophecy actually say, word for word? What does the hag think she is owed?

Every one of those questions comes up mid-scene, with five people waiting, and every one of them either gets answered fast or derails the night into a continuity hearing. The wiki that answers them is worth its weight in gold. The wiki that answers “what were the major exports of the eastern provinces” is a hobby. A lovely hobby. I have enjoyed that hobby. But it will not save your Tuesday.

This reframe sorts everything. An article earns its place in the wiki the first time somebody opens it during play. If you can’t imagine the mid-session question a page answers, the page is worldbuilding, and worldbuilding lives happily in a different document that no one has to maintain under deadline.

Why most campaign wikis die

I have watched a lot of wikis be born and I have watched most of them die, mine included, and they die on a schedule.

The encyclopedia dies first. This is the wiki written before the campaign: cosmology, history, factions, all of it authored in a burst of pre-game enthusiasm. It dies because none of it is load-bearing. The players never met these facts. Nothing at the table depends on them. When play generates real questions, the encyclopedia has no answers, because the questions are always about what happened at the table, and the encyclopedia was written before the table existed.

The homework wiki dies slower and sadder. This one starts honest: after each session, you write up what happened. Session one gets four paragraphs. Session five gets two. Session nine gets a bullet that says “fought guys, met Vessa?” and session twelve gets nothing, because the write-up is a forty-five minute job that lands exactly when you are most tired and least obligated to anyone. You know this arc if you’ve lived it. I wrote about it in You Shouldn’t Have to Take Notes, and the short version is that any system whose input is “the DM does unpaid clerical work after every session” has a half-life of about a month.

Both deaths have the same cause. The wiki was treated as a writing project, and the writing was pointed at the wrong subject.

Write what the table established. Nothing else.

Here is the rule that makes a wiki small enough to survive: the wiki records what was established at the table, out loud, with witnesses. Your plans are drafts. Your prep is a draft. The moment something is said in play, it hardens into canon, and canon is the only thing worth indexing.

This has a liberating corollary: the wiki starts empty. Session one produces maybe four entries. A tavern, two NPCs, one promise. That’s the whole wiki. It grows at exactly the rate the campaign does, which means it is never behind and never bloated, and every single page in it is something your players have actually touched.

What accumulates, concretely:

People. Name, one line of who they are, and their ledger with the party. “Vessa. Fence in Coppergate. Moved the candlesticks for 10%; believes Tam still owes her an introduction.” The ledger line is the part you will look up. Names on their own go stale, which is a thing I covered in the NPC names piece: a name attached to nothing is a stranger by Thursday.

Promises and debts. This deserves its own category, separate from people, because promises are the connective tissue of a long campaign and the single most disputed class of fact at any table. Who owes whom, what exactly was promised, and whether it has been paid. Half the arguments in your campaign’s future are sitting in this table waiting to be prevented.

Exact wording. Prophecies, contracts, riddles, the terms of the deal with the devil. Anything where the precise phrasing matters later. Paraphrase is poison here; the entire dramatic payload of a prophecy is in the wording, and six months from now nobody’s paraphrase will agree with anyone else’s.

Established facts about places. The wiki page for a city holds what happened there and what the party believes about it. The founding date can stay in your worldbuilding doc until the day a player asks, at the table, about the founding. Then it graduates.

The maintenance answer

So who writes it? Because even a lean wiki takes writing, and I just spent three paragraphs explaining why the DM doing it as homework fails.

If you take nothing else from this piece, take this: put the pen down at the table entirely and lower your standards everywhere else. One line per new entity, added whenever you next touch the wiki, is a wiki that survives. Fifteen entries with one honest line each will settle more disputes than three beautiful pages. I run my hand-written capture this way, on a five-line index card, and the card’s entries become wiki lines almost verbatim.

And this is the part of the problem we built Loracle for. The bot sits in your Discord voice channel, records the session, transcribes it, and builds the recap and the campaign wiki from what was actually said. The wiki that comes out is the one this whole piece has been describing, by construction: it only contains what happened at the table, because the recording is the source. The people your party met, the places they went, the exact wording of the deal, accruing session over session while you do the fun part of the job. Nobody does homework. The canon writes itself down.

However you produce it, hold your wiki to the ten-second test. Open it mid-scene, ask it who the smuggler is and what the barge deal was, and see if the scene survives. A wiki that passes is one of the best tools a long campaign can have. It ends arguments, it pays off six-month-old promises with the exact words, and it makes your world feel solid in the way only a world with receipts can feel.

The kingdom’s founding can wait. The barge cannot.