What sentence patterns does AI overuse?

The recurring ones are hedged assertions, three-item lists, em-dash appositions that define a term mid-sentence, the not-X-but-Y reversal, empty participial openers, self-referential transitions, false balance, nominalisation, uniform paragraph blocks, summary conclusions, unearned superlatives, and rhetorical questions as section openers.

Why does swapping words not fix AI writing?

Because the tell is structural. Replacing furthermore with additionally leaves the sentence shape, the paragraph rhythm, and the hedging exactly where they were. You can substitute every flagged word in a draft and it will still read as machine-made, which is why structural editing does far more than vocabulary editing.

What is the rule of three in AI writing?

The habit of producing three parallel items regardless of whether the subject has three parts — fast, reliable, and easy to use. Three has genuine rhetorical force, which is exactly why it gets overused into meaninglessness. Test it by removing one item: if the sentence loses nothing, the item existed to complete a pattern.

How do I spot machine-written prose?

Look at shape rather than words. List the first three words of every sentence in a section — repeated openers surface immediately. Count sentence lengths in one paragraph; clustering within a few words is a strong signal. Then check whether every paragraph is the same length, which is visible from ten feet away.

Are these patterns always wrong?

No. Every one appears in good human writing occasionally, and some are genuinely useful — an em-dash apposition can be elegant, and a tricolon can land. The signature is density and consistency: four or more per paragraph, applied uniformly across a page, with no variation.

How do I fix these patterns in my own writing?

Read it aloud, then delete before rewriting — the hedged openers, empty participials, signposting transitions, and summary conclusions mostly need removing rather than rewording. Then hunt repeated sentence openers, then demand a number in place of every intensifier.

12 Sentence Patterns That Make Prose Sound Machine-Made

A structural guide to the sentence shapes AI models overuse, each shown with a before-and-after rewrite and the reason the pattern exists.

Most advice about spotting AI writing is a list of banned words. "Delve." "Tapestry." "In today's fast-paced world." Those lists are useful — we have one of our own — but they are treating a symptom.

The deeper tell is structural. Language models produce a narrow range of sentence shapes, and they produce them over and over. You can swap every flagged word for a synonym and the writing will still read as machine-made, because the architecture underneath is unchanged.

What follows is twelve of those shapes, each with a rewrite and an explanation of why the model reaches for it. Learning to see them is worth more than any word list, because once you can see a shape you can fix it in writing that uses none of the obvious vocabulary.

1. The hedged assertion

Pattern: A claim wrapped in a phrase that announces the claim is coming.

It is important to note that regular maintenance can significantly extend the lifespan of your equipment.

Regular maintenance roughly doubles how long equipment lasts.

The opener contributes nothing. It exists because models are trained toward cautious, non-committal framing, and "it is important to note that" is the highest-frequency way English signals caution. Watch also for "it is worth mentioning," "it should be noted," "one must consider," and "it is essential to understand."

The fix is nearly always deletion. If the point is important, its importance shows in the fact that you wrote it down. Note that the rewrite also replaces "significantly" with an actual quantity — vague intensifiers are the same instinct in a different costume.

2. The tricolon

Pattern: Three parallel items, endlessly, regardless of whether the subject has three parts.

The platform is fast, reliable, and easy to use.

Our approach is grounded in research, refined through practice, and validated by results.

The platform is fast and it rarely breaks.

We built the approach from published research, then spent two years finding out which parts survived contact with real users.

Three-item lists have genuine rhetorical force, which is exactly why they get overused into meaninglessness. The giveaway is that the three items are frequently not distinct — "fast, reliable, and easy to use" mixes two performance claims with an unrelated usability claim, held together only by the rhythm.

The test: remove one item. If the sentence loses nothing, the item was there to complete a pattern rather than to say something. Cut to two, or expand one into a real sentence.

3. The em-dash apposition

Pattern: A noun, an em dash, a restatement of the same noun in different words.

Sentence rhythm — the variation in length and structure across consecutive sentences — is what makes prose feel alive.

Rhythm is what makes prose feel alive. It comes from varying sentence length so consecutive sentences don't all land the same way.

Models love this construction because it lets them define a term without committing to a full sentence. Used once, it is elegant. Used three times a page, it becomes a tic — and it produces uniformly long sentences, flattening the rhythm the writing is supposedly about.

We are not against em dashes; there are several in this article. The problem is specifically the definitional apposition, repeated.

4. The "not X, but Y" reversal

Pattern: Setting up a negation to make an ordinary claim feel like a revelation.

This isn't just about writing better — it's about thinking more clearly.

The problem isn't the technology. It's how we use it.

Writing clearly forces you to think clearly. That is most of the value.

The technology works. The way most teams deploy it doesn't.

This construction manufactures profundity. It implies a common misconception ("you probably think this is about X") and then corrects it, which flatters both writer and reader. The trouble is that the misconception is usually invented, and once you notice the shape you cannot stop noticing it.

Use it when there genuinely is a widespread wrong belief you are correcting. Otherwise, state the thing.

5. The empty participial opener

Pattern: A sentence beginning with a participial phrase that adds no information.

Building on this foundation, the team expanded into three new markets.

Considering these factors, it becomes clear that the strategy needs revision.

The team then expanded into three new markets.

The strategy needs revision.

These openers are connective tissue with nothing on either end. They perform the gesture of logical continuity without carrying an actual logical relationship. "Considering these factors, it becomes clear" is a particularly common one, and it is pure filler — either the factors make it clear, in which case say so, or they don't.

6. The self-referential transition

Pattern: Sentences that describe the structure of the text instead of advancing it.

Now that we've covered the basics, let's dive into the more advanced techniques.

In this section, we will explore three key strategies.

The advanced version of this is where it gets interesting.

Three strategies are worth the effort.

Signposting is a habit from formal instruction, and models absorbed a great deal of instructional prose. In practice, readers do not need to be told that a new section has started — the heading did that. Every one of these sentences is a sentence spent on navigation rather than content.

7. The false balance

Pattern: An acknowledgement of complexity that resolves into nothing.

While there are certainly benefits to this approach, it's important to consider the potential drawbacks as well.

This approach is faster to build and considerably harder to maintain. Whether that trade is worth it depends on how long the project will live.

The original sentence contains no information. It asserts that a thing has benefits and drawbacks, which is true of every thing. Models produce this constantly because balance is heavily reinforced in training, and generic balance is the cheapest way to appear balanced.

Real balance names the specific benefit, the specific cost, and the condition under which the trade makes sense.

8. Nominalisation

Pattern: Verbs converted into nouns, requiring a weak verb to prop them up.

The implementation of the new system resulted in an improvement in efficiency.

We conducted an analysis of the data and made a determination.

The new system made us more efficient.

We analysed the data and decided.

"Implementation," "improvement," "analysis," "determination" — these are all actions turned into things, which then need "resulted in," "conducted," and "made" to do the actual work. The result is longer, flatter, and vaguer, because a noun can hide who did what while a verb usually cannot.

Search your draft for -tion, -ment, -ance, and -ity endings. Each one is a candidate for turning back into a verb.

9. The uniform paragraph block

Pattern: Every paragraph three to four sentences long, every sentence fifteen to twenty words.

This one is invisible at sentence level and obvious from ten feet away. Squint at a page of generated prose and it looks like a wall of identical bricks.

Human writing is lumpy. A paragraph might be one sentence, because that sentence needed room. The next might run eight sentences because the idea kept going. That variation is not decoration — it is how emphasis works. A one-line paragraph after a long one lands.

Models produce uniform blocks because they optimise locally, sentence by sentence, with no sense of the shape of a page. The fix is structural: find your most important paragraph and cut it to a single sentence, and find two short adjacent paragraphs covering one idea and merge them.

10. The concluding restatement

Pattern: A final paragraph that summarises what was just said and adds nothing.

In conclusion, sentence variety is essential for engaging writing. By varying your sentence structure and avoiding repetitive patterns, you can create prose that resonates with readers.

The practical version: read your draft aloud. Where you run out of breath, cut. Where three sentences in a row start the same way, rewrite two of them.

Essay-format training runs deep, and the summary conclusion is its most durable artefact. But a reader who just read your article does not need it recapped, and a conclusion that only recaps is a paragraph of dead weight in the position of maximum emphasis.

End on the most useful thing you have, or on the implication your argument opens up. Not on a mirror.

11. The unearned superlative

Pattern: Intensifiers standing in for evidence.

This is a game-changing approach that will revolutionise how you work.

The results were incredibly impressive and highly significant.

This cut our review time from two days to about three hours.

Error rates dropped by a third.

"Game-changing," "revolutionary," "cutting-edge," "incredibly," "highly," "extremely" — each is a claim about magnitude with no magnitude attached. Models generate them because marketing copy is abundant in training data and superlatives are cheap.

Every intensifier is an invitation to replace it with a number, a comparison, or an example. If you cannot supply one, the claim probably was not true.

12. The rhetorical question opener

Pattern: Opening a section by asking the reader something.

Ever wondered why some writing just flows better than others?

What makes a photograph feel authentic?

Some writing flows and some doesn't, and the difference is usually rhythm rather than vocabulary.

Authenticity in a photograph comes down to texture — specifically, whether the noise behaves like a sensor's.

A rhetorical question defers the point by one sentence and asks the reader to do work before receiving anything. Occasionally that tension is worth creating. Most of the time it is a warm-up the writer needed and the reader does not.

Working through a real draft

Knowing the patterns is not the same as removing them. A practical order of operations:

Read it aloud. Nearly every pattern here is audible before it is visible. Filler openers make you stumble; uniform rhythm makes you drone.
Fix structure before words. Merging paragraphs and varying sentence length changes how a page reads more than any amount of synonym-swapping.
Hunt openers. List the first three words of every sentence in a section. Repeated openers will be obvious immediately, and they are the fastest thing to fix.
Delete before rewriting. Patterns 1, 5, 6, and 10 are usually solved by deletion alone. Do those first; the draft gets shorter and the remaining problems get easier to see.
Demand a number. Every intensifier and vague quantifier either gets replaced with something concrete or gets cut.
Read it aloud again. Editing introduces its own awkwardness.

Our Text Humanizer does a version of passes 2 and 3 automatically — it varies rhythm and restructures repeated openers while holding the meaning fixed. It will not do pass 5 for you, because supplying a real number requires knowing the number. That part stays yours, and it is usually where the biggest improvement is anyway.

For the follow-on work, the editing checklist covers the pass-by-pass workflow in more detail.

These twelve are structural symptoms. The habits that prevent them in the first place are in how to write like a human.