Crutch words are the words you reach for without deciding to: the fillers ("just," "very," "really"), the pet adverbs ("suddenly," "quietly"), and the signature content words you fell in love with once and never put down. Every author has a set. You cannot hear your own, because to you they sound like language, not like habit. The only honest way to find them is to count, and the count is usually a shock: you wrote this word 347 times.
One "shimmered" is a lovely word. The fortieth "shimmered" in the same book is a tic, and by then your reader has started noticing the word instead of the light. This close read covers what crutch words are, why a series multiplies them, why a naive frequency count tells you nothing, and how to turn an honest count into a revision pass, category by category. The rule underneath all of it: repetition you chose is style, and repetition you didn't choose is noise.
A crutch word is any word that appears in your prose far more often than you would ever choose on purpose. It is not a wrong word. It is a default: the word your drafting brain grabs when it is busy solving a harder problem, like the scene. That is why crutch words cluster in your best passages as much as your worst ones. When the writing is going well, you are not watching your word choice at all.
You cannot self-diagnose by reading. Your eye skips your own defaults the same way you stop hearing your own accent, and every individual instance looks defensible on the page, because individually, each one is. The eye reads "she just smiled" and registers a smile. It takes a count to reveal that "just" appears 412 times, which is once every 200 words, which means your reader meets it on every single page.
You can't hear your own crutch words. To you they sound like language. To a count, they look like a habit.
In a standalone novel, a crutch word is a local irritant. In a series it compounds, for two reasons that have nothing to do with your skill and everything to do with how series get written and read.
First, inheritance. Book 2 is written by the author Book 1 trained. Whatever tics you drafted into the first book, you carried into the second with interest, because by then they were not habits you fell into, they were "how this series sounds." A word you overused in Book 1 almost never corrects itself in Book 3. It settles in. The same drift that lets a narrator slowly stop sounding like themselves lets a pet word quietly become series furniture.
Second, binge reading. Series readers do not read your books a year apart the way you wrote them. They read Book 1 through Book 4 in a week. A word that appears forty times per book appears a hundred and sixty times in their week, and repetition that was subliminal at novel scale becomes conscious at series scale. This is the point where a reader starts a mental tally, and a reader keeping a tally is a reader who has left the story. If you write in a genre where box sets and binge reads are the norm, and if you are an indie series author you do, the crutch-word bar is simply higher for you than for a standalone novelist. The craft habit that meets it is auditing each book's word habits before it ships, so Book 3 does not inherit what Book 1 never fixed. The broader version of that discipline is the subject of the whole Series Craft pillar.
The obvious move is to count every word in the manuscript and sort by frequency. Try it, and here is the top of the list for any novel ever written: "the," "and," "to," "of," "a." Keep scrolling and the first interesting-looking entry is your protagonist's name, followed by the rest of your cast in order of page time. None of that is signal. Of course "the" is your most frequent word. Of course "Marta" appears two thousand times; she is in every scene. A raw count measures English and your cast list, not your habits.
To get signal, two exclusions have to happen before ranking, and the second one is the load-bearing one.
The name exclusion is why generic word-frequency tools disappoint novelists. A tool that does not know your world cannot tell "Ash" the character from "ash" the noun, or that "the Ember Crown" is a proper noun and not an overused pair of content words. The exclusion list has to come from your own story's canon, which is exactly the kind of structured knowledge a series author should be keeping anyway.
Raw counts also mislead across books of different lengths, and series are rarely uniform. Say "gripped" appears 90 times in your 60,000-word novella and 150 times in your 180,000-word epic. The raw counts say the epic has the problem. The rate says the opposite: the novella runs 15 per 10,000 words, the epic 8.3. Your reader experiences the rate, not the total, because the rate is how often the word interrupts a sitting.
So normalize: divide the count by the book's total words and multiply by 10,000. The resulting "per 10k" figure is the honest unit for comparing a word against the rest of your list, against the same word in another book of yours, and against your own sense of tolerable. As a loose calibration, a distinctive content word running past 8 per 10k is usually pressing on the reader's attention; a filler can run far higher before anyone consciously notices, which is precisely what makes fillers so durable.
Your reader experiences the rate, not the total. Per 10,000 words is the honest unit, and it is what lets a novella and an epic look each other in the eye.
Once the list is clean and normalized, do not attack it top to bottom. Different kinds of overused words fail differently and deserve different revision moves. Three categories cover nearly everything that survives the filters.
"Just," "very," "really," "quite," "rather," "somewhat," "actually," "basically," "literally," "suddenly," "simply," "perhaps," and the perception hedges "seemed" and "felt." These are the classic prose-tightening targets, and the fix is almost always deletion. Read the sentence without the filler; in nine cases out of ten it is the same sentence, sharper. "Seemed" and "felt" earn a second look, because they often mark a spot where you hedged instead of committed: "the room seemed cold" is a weaker choice than the room simply being cold, unless the uncertainty is the point.
The -ly words are not automatically sins, whatever the internet says, but each one is a question: is this adverb propping up a verb that could stand on its own? "Walked quickly" is "strode." "Said quietly" is "murmured." "Held tightly" is "gripped." When the -ly section of your list is long, it usually means your verbs are doing half a job each. Keep the adverbs that genuinely change meaning; trade the rest for stronger verbs, one at a time, by hand.
This is the "shimmered" case: a vivid, distinctive word you love, and that is the problem. Distinctive words spend the reader's attention every time they appear, so their budget is small. A word like "shimmered" might earn five appearances in a novel; at 347 it has stopped meaning anything. Do not purge it. Ration it: keep the handful of instances where it is exactly right, and revise the rest into varied, specific alternatives. The goal is that when the word does appear, it lands like a choice again.
Here is the pass, whether you run the counts by hand or have a tool do the counting for you. It slots naturally between the developmental edit and the line edit: late enough that scenes are stable, early enough that you are still willing to touch every page.
Count every word in the finished draft, exclude stopwords and every name in your story, and rank what remains with a per-10k rate next to each count. If you are doing this manually, a word processor's find-all count against a shortlist of suspects is a workable, if partial, substitute.
Fillers, -ly adverbs, content words. The category tells you the move before you open the manuscript: cut, audit, ration. Mark the two or three content words at the top; those are your signature tics, and they will be the highest-value fixes.
Do a filler pass, then an adverb pass, then one pass per rationed content word. Category passes keep your judgment consistent; page-order passes let your tolerance drift as you tire. And every change is a decision, not a rule: if "just" is doing real work in a line of dialogue, it stays.
After revision, count again and confirm the rates actually moved. Then keep the list. Your crutch words are stable across books, and knowing that "shimmered" was Book 1's tic is the cheapest possible insurance against it becoming the series' tic. Checking each new book against the last book's list is how a series stays tight instead of accumulating habits.
StoryHelm's Style lens runs exactly this count over your manuscript. It is deterministic counting over your actual prose, not an impression: every word in the book is counted at word boundaries, stopwords are excluded, and, because your characters and places already live in your Canon, every name and alias in your story is excluded too. That canon-aware exclusion is what keeps your protagonist off the top of the list and lets the real habits through.
What comes back is a headline, "Your most overused word: shimmered, 347 times," and a ranked list beneath it, each word with its exact count, its per-10k rate, and its category chip: filler, adverb, or content. The counts are real counts from your own book, so "347" means 347, and the rate means you can compare this book's habits against any other length of book. A finding might take roughly this shape:
Write your series in StoryHelm's Scene Editor or import books you finished elsewhere; your prose, your Canon, and the analysis all live in one place. The Style lens counts every word in the book you point it at, excludes stopwords and the character and place names from your Canon, and hands you the headline word plus the ranked, categorized list with exact counts and per-10k rates. Run it on each book in your series and set the lists side by side to see which habits repeat. It shows you the list; you decide what stays. StoryHelm reads and analyzes; it never rewrites a word of your prose. For the companion habit on the voice side, see character voice drift, and for the filler-density metric inside Plot DNA, see filler ratio.
No. Deliberate repetition is a tool: a motif, a refrain, a character's verbal tic you planted on purpose. The test is authorship. If you can say why the word recurs and where, it is style. If the count surprises you, it was a habit, and habits are what this pass exists to catch. The list is information, not a verdict; plenty of words on it will survive your review, and should.
There is no universal number, which is why the per-10k rate matters more than the raw count. Structural fillers can run high before anyone notices. A distinctive content word is on a much tighter budget; past roughly 8 per 10,000 words it starts spending reader attention every time it appears. Genre moves the line too: a lyrical fantasy tolerates more "shimmered" than a procedural. Calibrate against your own list, not someone else's rulebook.
Because names are frequent for structural reasons, not stylistic ones. Your protagonist appears in nearly every scene, so her name will out-count any genuine crutch word by an order of magnitude, and every generic frequency tool that doesn't know your story will put her at the top of your "overused" list. Excluding the names and terms from your own canon is what turns a word count into a craft instrument.
No. StoryHelm shows you the word, the exact count, the rate, and the category, and the revision is yours. Whether the fortieth "shimmered" stays or goes is a judgment about your book that only you can make. StoryHelm reads and analyzes your prose; it never writes or rewrites a word of it.
The word-habit problem's bigger sibling: when a narrator's diction, rhythm, and tics slowly change across a series until they stop sounding like themselves.
ReadThe full tour of the craft dimensions a series is judged on, with a close read for each of the ways long stories quietly come apart.
ReadStoryHelm reads your whole series at once, pairs every setup with its payoff, and hands you the list of threads that pay off nowhere, pointed to the exact book and chapter where the promise was made. Catch the dangler now, not in the one-star review from the reader who carried it through five books waiting for the answer.
Get the free Self-Audit