The Adverb Diet: Cutting ‘-ly’ Words to Strengthen Verbs and Sound Human

Open most rough drafts and you’ll find the same quiet habit: a verb that does half its job, propped up by an adverb doing the other half. “Walked slowly.” “Said angrily.” “Significantly improved.” Each one feels harmless. Stacked across a page, they drain the life out of your sentences and, increasingly, they make your writing read like a machine wrote it.

This isn’t a grammar lecture. It’s a practical edit you can run on any draft in ten minutes, and it happens to fix two problems at once: weak prose and an AI-flavored texture that both detectors and real readers notice.

Key takeaways

• Most ‘-ly’ adverbs are patching a weak verb. Pick a stronger verb and the adverb falls away.
• Adverbs aren’t the enemy. The ones to cut are the ones that repeat the verb or hedge a clear claim.
• Heavy adverb use is a common AI tell, because models lean on safe intensifiers like “significantly” and “effectively.”
• The method is a question, not a rule: does the sentence lose meaning without this word? If not, cut it.
• Stronger verbs add rhythm and specificity, which is a big part of what makes writing sound human.

What an adverb is actually doing in your sentence

Grab any ‘-ly’ adverb and ask what it’s there for. Usually it’s compensating. “He ran quickly” uses two words to say what “he sprinted” or “he bolted” says in one, with more punch. The adverb is a crutch for a verb that wasn’t pulling its weight.

That’s the core insight behind why you should cut adverbs in writing: the adverb isn’t the disease, it’s the symptom. It points to a verb you can upgrade. When you swap in the stronger verb, the adverb has no job left and quietly exits.

Compare these:

• *She quickly closed the laptop and walked nervously to the door.*
• *She snapped the laptop shut and hurried to the door.*

Same scene. The second is shorter, more specific, and it moves. “Snapped” and “hurried” carry the speed and the nerves that “quickly” and “nervously” were trying to bolt on from the outside.

Not all adverbs are guilty

Here’s the part the strict “delete all adverbs” crowd gets wrong. Some adverbs change the meaning, and those stay.

“She nearly finished the report” is not the same as “she finished the report.” “Nearly” is doing essential work. Same with “he only paid for one ticket” versus “he paid for one ticket” — the “only” sets up an expectation. Adverbs that add information, contradict, limit, or surprise are doing their jobs. Keep them.

The diet targets a narrower set: adverbs that repeat the verb (whispered quietly, screamed loudly, gently caressed) and adverbs that hedge a statement into vagueness (this is really quite important, the results were somewhat significant). Those are the empty calories.

Why adverb pileups read as AI

Language models are trained to predict the next likely word, and likely words are safe, common, middle-of-the-road words. That tendency shows up in word choice. When an AI describes something as improved, it often improves it “significantly.” When it explains a method, the method works “effectively” or “seamlessly.” These intensifiers are everywhere in training data, so they’re everywhere in the output.

The result is a texture problem. The writing gets evenly hedged, smoothly qualified, and a little weightless — every claim padded with a soft adverb so nothing lands too hard. Detectors don’t flag a single “significantly.” They respond to the overall statistical evenness of a passage, and adverb-padding flattens that texture. If you want the mechanics of how that measurement works, we walk through the statistics behind how detectors read your writing in a separate piece.

Human writers are messier and more committed. We pick a vivid verb and let it stand. We don’t write “the deadline rapidly approached” — we write “the deadline loomed.” That decisiveness is part of what reads as a real person behind the keyboard.

A quick before-and-after

Take a sentence that smells like an AI draft:

> *The new policy will significantly help teams to effectively collaborate and ultimately improve productivity meaningfully.*

Count the adverbs: significantly, effectively, ultimately, meaningfully. Four hedges in nineteen words, and somehow it says almost nothing. Strip them and force the verbs to carry the load:

> *The new policy helps teams collaborate and ship faster.*

Eight words. It commits to a claim. You could argue with it, which is a good sign — vague writing is unarguable because it never actually says anything.

The five-minute adverb diet

Here’s the pass, step by step. Do it after your draft is done, not while you’re writing. Drafting and editing are different jobs.

1. Find them. Search your document for “ly ” (with a trailing space). Most filler adverbs end in -ly, so this catches the bulk of them in seconds.
2. Read each one and ask: does the sentence lose meaning without this word? Say the sentence aloud both ways.
3. If it loses nothing, delete it. “He completely finished” becomes “he finished.” Done.
4. If the sentence feels weak after deleting, that’s the signal to upgrade the verb, not to put the adverb back. “Walked slowly” with the adverb gone is just “walked” — flat. So reach for “shuffled,” “ambled,” “trudged,” whichever fits.
5. Keep the ones that change meaning. Nearly, only, almost, barely, hardly — these usually earn their keep. Don’t be dogmatic.

Run this on a 1,000-word draft and, in my experience, you’ll often cut a couple dozen words and strengthen a handful of verbs along the way. The piece gets shorter and sharper at the same time.

A mini-scenario

A student sends me a paragraph from a lab report:

> *The temperature gradually rose, and the solution slowly changed color, eventually turning completely clear. We carefully recorded the data and then quickly moved to the next trial.*

We run the diet together. “Gradually rose” → “climbed.” “Slowly changed color, eventually turning completely clear” → “drifted from cloudy to clear.” “Carefully recorded” stays, because in a lab “carefully” is a real claim about method. “Quickly moved” → “moved to the next trial” (the speed wasn’t the point).

> *The temperature climbed, and the solution drifted from cloudy to clear. We recorded the data carefully and moved to the next trial.*

It still reports the same facts. It just stopped narrating in slow motion. And — not coincidentally — it stopped sounding like it came out of a chatbot, because the verbs now do specific work instead of leaning on generic speed-and-degree padding.

Where this fits in the bigger picture

The adverb diet is one habit, not a complete style. It pairs with a handful of others: varying your sentence length, choosing concrete nouns, cutting the throat-clearing phrases (“it’s worth noting that,” “in order to”). Together these are what give writing the unevenness and specificity that reads as human. You can browse more writing-craft articles if you want the rest of the toolkit.

It’s also worth being honest about what this buys you. Cleaner verbs make a draft read better, full stop. They can also nudge a detector’s score, but a detector only ever returns a probability, not a verdict — and those tools have a documented track record of getting it wrong. OpenAI pulled its own AI-text classifier in July 2023 for low accuracy, and a 2023 Stanford study (Liang et al.) found several detectors flagged writing by non-native English speakers as AI far more often than native writing. So treat the diet as craft first and a detection tweak second.

One more caution: this is editing, not a trick. Don’t replace adverbs mechanically with a thesaurus, swapping “very” for “exceedingly” and calling it done. That keeps the same weak structure and just dresses it up. The point is to rebuild the verb, which often reshapes the whole sentence — and that deeper rework is what actually moves the needle on how your writing reads.

Frequently asked questions

Should I delete every adverb in my writing?

No. Cut the lazy ones, keep the working ones. Adverbs that change meaning — “she nearly finished,” “he only paid once” — earn their place. Hunt down the ones that just echo the verb (“whispered quietly”) or hedge a clear point into mush (“really quite important”). Read each ‘-ly’ word and keep the ones doing real work.

Why do adverbs make writing sound AI-generated?

Models reach for safe, high-frequency words, and generic intensifiers like “significantly,” “effectively,” and “seamlessly” fit that pattern. A draft studded with them gets a flat, evenly-hedged texture that both detectors and careful readers pick up on. It’s not any single adverb — it’s the predictable density of them.

How is cutting adverbs different from using a synonym tool?

A synonym swapper trades one word for a near-match and leaves the sentence shape untouched, which is shallow and easy to spot. Cutting an adverb forces you to rebuild the verb, which often reshapes the whole sentence. That’s a real editorial decision, not a find-and-replace.

Will trimming adverbs help me pass an AI detector?

It helps, but it’s one ingredient. Detectors output a probability, not proof, and they respond to the overall texture of your writing — how varied your word choices and sentence lengths are. Strong verbs and fewer filler adverbs add that variety. Pair the adverb diet with varied sentence length, concrete detail, and your own voice, and the whole draft reads more naturally.

Put it to work

Try the diet on your next draft, then check whether it actually moved the needle. Paste the before and after into a detector, watch which sentences cool off, and you’ll start to feel which verbs were carrying weight and which were faking it. When you want a faster read on where a draft still sounds machine-made, run it through PaperBleach and let the heatmap show you the spots your verbs haven’t earned yet.