The same glossary has to serve a reader starting book one, a reader waiting for book five, and you. Those are three different documents with three different cut points, and the way to have all three without maintaining three is to keep one master where every entry knows which book it belongs to.
The levels
| Level | Contains | Used for |
|---|---|---|
| Working master | Everything, including unpublished plans | You, and nobody else |
| Current to latest book | Everything published so far | The website, and the newest book's back matter |
| Current to book N | Cut at that book's end | Each individual book's back matter |
| Spoiler-free | Terms and neutral definitions only | Front matter, samples, marketing |
| Reader-facing pronunciation | Names and sounds, no facts | Audiobook listeners, and anyone reading aloud |
The tagging scheme
- Every entry gets a first-appearance book number. This alone produces the cut-at-book-N versions automatically.
- Every entry gets a spoiler flag, for the entry as a whole.
- Entries with staged information get one definition per stage, because a character's meaning changes across a series.
- Filter, don't copy. A published version is a filtered view of the master, generated when you need it.
- Record which version went where, so you know what a reader of book two has already been told.
Point three is the subtle one. An entry reading the captain of the guard is safe until book three, when the captain turns out to be someone else. The entry does not become wrong; it becomes a spoiler if worded with hindsight, and it needs a book-two wording and a book-four wording.
The wording trap
Most spoiler leaks in glossaries are not facts, they are tenses and adjectives. The former queen tells a reader she stops being queen. Amara's first husband tells them there is a second. Write each entry at the level of knowledge the reader has reached, and read the whole filtered version once before publishing it, looking specifically at word choice rather than content.
Where the master list comes from
The Novel Kit extracts the terms from each manuscript with first appearances, which supplies the book-number tag for every entry as a by-product. Run it per book and the tags are evidence rather than recollection, which matters because remembering which book introduced a term is exactly the thing that fails at book four.
It does not tag spoilers or judge wording, and the filtered views of a series-wide master are yours to generate; its own spoiler-level editions cut a single book's glossary at a chosen point. It gives you the term, the count and where it first occurs; the spoiler judgement is editorial and it stays with you.
A practical next step
Open your current glossary and add a book number to every entry. That one column gives you every cut-at-book-N version immediately, and it's the change that makes the rest of this practical.
Sources checked
Frequently asked questions
How many versions do I actually need?
Usually three: the working master, a version current to the latest book, and a spoiler-free front-of-book-one version.
Can I not just write three documents?
You can, and they will disagree within a year. Tagging one document and filtering it is the only version that stays true.
What counts as a spoiler in a glossary?
A death, a revealed identity, a betrayal, and any definition whose wording implies a later event. The wording is the part people miss.