If your series has more than one invented name in it, you already have a consistency problem waiting to happen. Glossaries don't just help readers, they're the record that keeps you honest across a hundred thousand words and a two-year drafting process.
Start with a first pass, not a final draft
Don't try to write final definitions on your first scan. Run the manuscript through a term extractor, let it surface every candidate (capitalised names, bolded phrases, invented technology), and treat the result as a long, messy shortlist. Trimming a list is much faster than building one from a blank page.
Watch for near-duplicate names
The most common glossary bug isn't a missing term, it's two versions of the same one. "Fadin" in chapter two and "Faddin" in chapter nineteen will both look correct to you on a normal read-through, because your brain autocorrects for meaning, not spelling. A fuzzy-matching duplicate check catches these long before an editor, or a reader, has to.
A glossary is only as trustworthy as its spelling is consistent. One typo in an appendix undermines the whole reference.
Let definitions surface from your own text
Chances are you've already defined half your invented terms in-line: "the Jump propulsion unit, a drive built for atmosphere-to-orbit hops" reads naturally in prose and doubles as a definition. Tools that scan for definition-like sentences can pull these out automatically, giving you a first draft of your appendix instead of a blank page. Always sense-check what comes back; it's a starting point, not a finished entry.
Export early, refine often
Keep your glossary as a living export: .csv while you're still cutting entries, .docx once you're ready to paste it into the back matter. Re-run the scan after major revisions rather than trying to keep the list updated by memory. The manuscript is the source of truth; the glossary should always be regenerated from it, not the other way around.
Carrying it from book one into book two
Everything above is a single-manuscript job. A series is that job repeated, plus one step that only exists once there is more than one book: setting the new list against the old one. Each book is scanned on its own, and a scan of book three knows nothing about book one, so the continuity has to come from comparing the exports rather than from the scan itself. That's the honest limit, and working with it is straightforward.
- Scan each book separately, on the same detection settings. Change a setting between books and the two lists are no longer comparable, which is the one mistake that wastes the whole exercise.
- Set the new book against the previous one. Compare two drafts takes the earlier manuscript, or a saved JSON export of it, and reports which terms arrived, which are no longer present, and which pairs look like a rename rather than one loss and one addition. Point it at book one and book two and the "no longer present" column is the list of things the sequel quietly stopped using.
- Read the arrivals first. A term that appears for the first time in book two is either genuinely new, in which case it needs a definition, or it is a drifted spelling of something from book one, in which case the pair usually shows up as a probable rename.
- Keep one master glossary and regenerate it, never edit it by hand. The master list is the union of the per-book exports. Hand-editing it is how a series glossary starts disagreeing with the books it describes.
- Bind the reference before you start the next book. The series bible generator puts the glossary, the people, the places and the consistency findings from a scan you have already run into one document, which is what you want open on the day you begin book four.
The drift to watch for in a series is the one with a gap in it: a name used constantly in book one, absent from book two, and then misspelled in book three because two years have passed. Within a single manuscript the fuzzy check catches that pairing on its own. Across books, only the comparison will, which is the argument for running it every time rather than only when something feels wrong.
None of this replaces your own judgement about what belongs in the final appendix. But it turns a weekend of highlighter-and-notepad work into something you can finish over a lunch break, leaving the actual writing time for the next book.
Frequently asked questions
How often should I re-scan my manuscript for glossary terms?
After every major revision. Regenerating the glossary from your latest draft beats trying to hand-update a list as terms are added or cut.
What causes near-duplicate character names?
Your brain reads for meaning, not spelling, so two versions of the same invented name can both look correct across a long manuscript.
Should I write my own glossary definitions or use suggested ones?
Suggested definitions pulled from your own text are a useful starting point, but always sense-check them before publishing.