How to find and fix the clichés hiding in your writing

Read your own draft out loud and most of the clichés will sail right past you. That is the whole trouble with a cliché: it is a phrase so worn that the brain stops registering it as words at all. "At the end of the day", "low-hanging fruit", "a perfect storm" — they slide in while you are concentrating on the idea, and because they feel like fluent writing, nothing inside you flags them. You need an outside pass to catch what your ear has gone deaf to.

That is exactly the job a cliché detector does: it reads your text against a curated list of roughly 150 well-worn phrases and points at every one it finds. This piece explains why clichés hurt your writing in the first place, what the six categories the tool hunts for actually are, how the matching works (and where it deliberately stops), and how to fix each hit rather than just delete it.

Why a dead phrase costs you

A cliché was a good image once. "The tip of the iceberg" was vivid the first thousand times; now it is furniture. George Orwell put the problem plainly in "Politics and the English Language": stale metaphors let a writer skip the work of thinking, gluing ready-made phrases together "like the sections of a prefabricated henhouse". The reader pays for that shortcut. A cliché signals generic — it tells the reader you reached for the nearest phrase instead of the precise one, and it makes confident writing read as filler.

The fix is rarely deletion. It is replacement with something concrete. "We picked the low-hanging fruit" becomes "we fixed the three bugs we could close in an afternoon". The cliché compressed a real thought into a stock image; your edit unpacks it back into specifics. That is why a detector is a starting point, not an autocorrect: it finds the phrase, but only you know what you actually meant.

The six families a detector hunts

The tool tags every match into one of six categories, because clichés cluster by where they breed:

• Business jargon — the largest group, around fifty phrases: "circle back", "move the needle", "synergy", "paradigm shift", "deep dive", "best practices", "going forward". These are the dead language of the meeting room, and they migrate into writing fastest.
• Idioms — the everyday set-phrases like "the tip of the iceberg" or "the bottom line" that pass for color but add none.
• Dead metaphors — images so faded they no longer summon a picture: "a perfect storm", "the elephant in the room", "level the playing field".
• Weasel phrases — the hedges that let a sentence imply something without committing to it: "studies show", "it goes without saying", "needless to say".
• Romance — the stock phrases of love-writing that turn fresh feeling into greeting-card copy.
• Sports — "hit it out of the park", "move the goalposts", "a slam dunk", "the home stretch", the metaphors that leak from the field into everything else.

You can switch any category on or off. If you are editing a deck where some business shorthand is genuinely expected, mute that family and let the tool concentrate on the idioms and dead metaphors you do want gone. The phrase list itself is drawn from public-domain sources — Plain Language guidelines, Garner's Modern English Usage, and Orwell's essay — so it reflects the phrases editors actually flag, not a random grab bag.

How the matching works — and where it stops

The detection is deliberately strict: each phrase is matched exactly, case-insensitively, with word boundaries. That last detail matters. "Tip of the iceberg" matches whether you wrote it capitalised or not, but "iceberg tip" does not — the tool is looking for the set phrase as it actually circulates, not any sentence that happens to mention an iceberg. Word boundaries also mean a cliché buried inside a longer word will not produce a false hit.

The trade-off is that the detector will not catch a cliché you have reworded. If you write "circling all the way back around", the canonical "circle back" entry will not fire. This is a feature, not a bug: it keeps false positives near zero, so every flag you see is a real, recognisable cliché rather than a guess. The cost is that the tool measures known clichés, not novelty in general — it cannot tell you a fresh phrase has quietly become your personal crutch. For that, you still need your own ear.

Because the list is editable, you can extend it. Drop your own pet phrases into the custom list — the in-house buzzword your team overuses, the transition you lean on in every post — and the detector will hunt those alongside the built-in set. That turns a generic checker into a personal one.

Reading the output: count, density, and category

The tool gives you three numbers worth reading together. The raw total is how many clichés it found. The per-100-words density is the more honest figure, because it is comparable across documents of different lengths — three clichés in a tweet is a crisis; three in a 2,000-word essay is a light edit. And the per-category breakdown tells you what kind of writer you are being on this draft: a wall of business-jargon hits means you have slipped into deck-speak, while a pile of dead metaphors means you are reaching for stock images instead of specific ones.

A practical workflow: run the draft, sort the flagged phrases by frequency, and start with whichever phrase repeats most. Repetition is the loudest signal — a cliché you used once is a slip, but one you used four times is a habit worth breaking. Replace each with the concrete thing it was standing in for, then re-run to confirm the density dropped.

Where this fits in an edit

De-clichéing is one pass in a larger self-edit, and it pairs naturally with the rest of the toolkit — trimming the hedges a filler word detector surfaces, or checking that your readability has not crept up while you were swapping phrases. Do the cliché pass after the draft is structurally done but before the final polish, when you can still afford to rewrite a sentence rather than just patch a word.

When you are ready to see what has gone stale in your own prose, paste it into the cliché detector. It will list every flagged phrase, tag it by family, show you the density per hundred words, and let you mute the categories you do not care about and add the ones you do — so the only thing left for you to do is the part no tool can: say what you actually meant.