AI Writing Tropes: The Sentence Shapes That Give It Away
What Are AI Writing Tropes?
AI writing tropes are the sentence shapes a model reaches for by default, the kind you can spot even after every flagged word is gone.
Vocabulary is easy to fix. A client swaps “leverage” for “use,” deletes an em dash, and hands the draft back to me looking clean.
Then I read one paragraph out loud and catch three sentence shapes that have nothing to do with word choice at all.
That’s the gap this piece covers. I already wrote up the fourteen signs of AI writing that span vocabulary, rhythm, and whole-page patterns.
This one zooms into a single level of that list, the shape of an individual sentence, and names nine tropes the wider list only mentions in passing.
Why Does Sentence Structure Matter More Than Word Choice?
Sentence structure matters more because a model can rewrite its vocabulary in a second, and a person will still feel the shape underneath.
A client once ran her homepage through three separate rewrite passes, swapping every flagged word each time.
The page still read like a template, because the whether-or hedge in her hero line never moved.
That’s the practical difference between an AI writing pattern at the word level and one at the sentence level.
Word-level patterns die the moment someone notices them. Sentence-level ones survive every edit that doesn’t touch the grammar.
What’s the Difference Between an AI Writing Trope and an AI Writing Sign?
A sign can be a word, a rhythm, or something you only catch across a whole page.
A trope is narrower. It’s one specific sentence shape, repeatable across any topic, that a model reaches for the same way every time.
The em dash is a sign. The trope underneath it is more precise: two half-ideas welded together where a full stop would’ve worked better.
I go deep on that particular trope, plus five others like it, in how to make AI writing sound more human.
This post picks up where that one leaves off. Nine shapes below don’t get their own section anywhere else on this site yet.
What Is Parataxis, and Why Does AI Keep Doing It?
Parataxis means placing clauses or sentences side by side with nothing connecting them, so each fact sits alone instead of building on the one before it.
“The client had a slow website. Traffic was flat. Conversions were low.” Three facts, no mortar between them.
A person reading it can feel the gap before they can name it, the same way you can hear a stranger reading off a list rather than talking to you.
A model defaults to this shape because predicting the next likely sentence doesn’t require linking it back to the last one.
A person writing from a real argument connects the two anyway, because the second thought grew out of the first.
I already broke down that exact pattern, sentence stacking, in the sentence-rhythm breakdown. I won’t repeat the whole rewrite here.
What follows are nine sentence shapes that show up once a draft has cleared the obvious stuff.
What Does Negative Parallelism Look Like in AI Writing?
Negative parallelism stacks two or three negated claims in a row before landing on the actual point, and it’s the single most common trope I flag.
Flagged, from a wedding photographer’s homepage: “Not the venue. Not the flowers. The two people standing in front of everyone they love.”
Two throwaway negations sit in front of the sentence that mattered most. Cut the setup, and the real line still lands.
What I sent back: “The photo that matters is the two people standing in front of everyone they love, not the venue or the flowers behind them.”
Same idea, one sentence instead of three, and it reads like a person said it once instead of building up to it.
What Is the “False Range” Sentence AI Keeps Reaching For?
A false range spans two extremes, “from confused to confident,” to imply a transformation without naming a single thing that happens in between.
Flagged, from a fitness coach’s sales page: “This program takes you from exhausted and inconsistent to strong and unstoppable in twelve weeks.”
That sentence describes a feeling at each end and skips the entire middle. The middle is where the coaching happens.
What I sent back: “This program gets you to the gym three mornings a week for twelve weeks straight, and week four is when most clients stop needing me to remind them.”
One real mechanic beats two adjectives at either end of an arrow.
Why Does AI Love the “Whether X or Y” Construction?
A model reaches for “whether you’re X or Y” because it sounds inclusive while committing to nothing, covering both audiences instead of choosing one to write for.
Flagged, from a SaaS onboarding email: “Whether you’re a solo founder or running a team of fifty, this dashboard scales with you.”
That sentence aims at everyone, and it lands soft because of it. A solo founder and a fifty-person team don’t have the same problem on day one.
What I sent back: “If you’re running the whole company from your kitchen table, this dashboard is built for you first. The team-of-fifty version can wait.”
Picking one reader over hedging toward both is what makes a sentence sound like it was written for someone.
What Is Anaphora Stacking, and How Do You Catch It?
Anaphora stacking repeats the same opening word or phrase across three consecutive sentences, borrowing the rhythm of a speech without earning it.
Flagged, from an agency landing page: “This is a system built for growth. This is a system built for clarity. This is a system built for people done guessing.”
It reads like a rehearsed shape rather than three separate thoughts, because a person doesn’t usually repeat their own sentence structure on purpose three times running.
What I sent back: “This system replaces guessing with a plan you can see coming.”
One sentence, said once, does the same job the triple repetition was straining for.
Want this done for you? Book a free strategy call →
Why Do AI Sentences Open With “-ing” So Often?
A model fronts a sentence with a gerund clause, “balancing the demands of the modern workplace,” before it’s told you who’s doing the balancing.
Flagged, from a B2B blog post: “Balancing multiple priorities while chasing growth, small business owners often struggle to keep marketing consistent.”
The subject shows up eleven words late, after a clause that could apply to almost any business owner in any industry.
What I sent back: “Small business owners chase growth and let marketing slip, because there’s only one of them and twelve priorities pulling at once.”
Put the subject first, and the sentence sounds like it’s about a specific person instead of a category of person.
What Is a Comma-Clipped Trailer?
A comma-clipped trailer tacks a vague modifying phrase onto the end of an already-finished sentence, adding rhythm without adding a fact.
Flagged, from a nonprofit newsletter: “The program served two hundred families this year, continuing to make a meaningful difference in the community.”
The sentence was complete at “this year.” Everything after the comma restates the same feeling in fewer specifics than the number already carried.
What I sent back: “The program served two hundred families this year, and forty of them came back to volunteer once they’d gotten back on their feet.”
Same comma, same position in the sentence, and this time it’s carrying a fact instead of a mood.
What Is Parallel List Overload?
Parallel list overload forces every item in a list into the same grammatical shape, even when the real items behind them aren’t that symmetric.
Flagged, from a consultant’s service page: “We help you plan your growth, execute your strategy, and measure your results.”
Three verb-noun pairs, evenly weighted, evenly spaced, describing three phases of work that in real life take wildly different amounts of time and attention.
What I sent back: “Most of my time goes into the strategy. Planning takes a week. Measuring takes an afternoon a month, once the dashboard’s built.”
A real list reflects how the work splits in practice, and the work almost never splits into three matching pieces.
Why Does AI Overuse Formal Transition Words?
A model reaches for a formal adverbial transition, “additionally,” “in addition,” in place of the plain conjunctions a person says out loud.
Flagged, from a course sales page: “The course covers pricing strategy. Additionally, it addresses client communication. In addition, students receive lifetime access.”
Three sentences, three separate facts, and a connector between each one that belongs in a printed report, not a conversation.
What I sent back: “The course covers pricing and client communication, and it comes with lifetime access, so you can rewatch the pricing module the week you need it.”
And, and, so. That’s the whole connective toolkit a person needs, and it’s the one AI reaches for last.
Is a Semicolon a Sign of AI Writing?
A semicolon by itself isn’t a tell. Plenty of careful human writers use one correctly on any given page.
What gives AI away is a semicolon joining two clauses of matched length and weight, the sentence equivalent of a matched pair of socks.
Flagged, from an agency about page: “Our team moves fast; our clients see results quickly.”
Both halves run the same length, make the same claim from two angles, and add up to one idea dressed as two.
What I sent back: “Our team moves fast, so clients usually see results inside the first month, not the first quarter.”
A real conjunction instead of a semicolon forces the sentence to say how the two halves connect, not just that they’re next to each other.
What Does an AI Sentence Look Like, Start to Finish?
Stack a few of these shapes in one paragraph and you get something like this, pulled together from the kind of homepage draft that lands in my inbox weekly.
Flagged: “Whether you’re scaling a startup or steadying a legacy brand, we walk beside you every step of the way. From confused to confident, that’s the whole journey. Juggling the constant churn of marketing, businesses need a partner they can trust; in addition, we deliver results.”
Three sentences, four tropes: a whether-or hedge, a false range, a gerund opener, and a matched-clause semicolon propped up by an adverbial transition instead of a real conjunction.
What I sent back: “I’ve worked with three-person startups and forty-year-old family brands, and the job stays the same either way: fewer guesses, more of what moves the needle. Most clients see a real shift by month two, once someone finally looks at their numbers and tells them the truth instead of handing them a template.”
Same claim, and every word of it is doing something.
9 AI Writing Tropes at a Glance
| Trope | What it looks like | Say instead |
|---|---|---|
| Negative parallelism | ”Not X. Not Y. About Z.” | Say Z, once |
| The false range | ”From overwhelmed to unstoppable” | Name the real mechanism in between |
| The inclusive hedge | ”Whether you’re X or Y…” | Pick the one you’re writing for |
| Anaphora stacking | The same opening words, sentence after sentence | Say it once, then move on |
| Dangling participle opener | ”Balancing X, businesses must…” | Put the subject first |
| Comma-clipped trailer | A vague phrase tacked on after a comma | Add a real fact, or cut it |
| Parallel list overload | Every list item forced into the same grammatical shape | Let the list match the real work |
| Adverbial overload | ”Additionally,” “In addition,” stacked as connectors | And, but, so |
| Semicolon symmetry | Two matched clauses joined by a semicolon | A period, or a real conjunction |
Are These the Same as the “Patterns of AI” Lists You See Elsewhere?
Not exactly, since every list of this kind draws its own lines around the same handful of habits.
My own count has moved around on purpose. The vocabulary graveyard covers 17 words.
The fourteen signs of AI writing span four levels from vocabulary to whole-page rhythm, and the sentence-rhythm breakdown narrows to six patterns you can fix without a full rewrite.
This page adds nine more that sit specifically at the level of one sentence’s shape, distinct from the six already covered and distinct from the fourteen broader signs.
Pick whichever list matches the problem in front of you. A single flagged word needs the graveyard. A whole draft that feels off needs the fourteen signs. One sentence that reads wrong needs this page.
How Do You Tell If a Sentence Was Written by AI?
Check the shape before you check the words, because a shape survives a synonym swap and a word list doesn’t.
Read the sentence out loud and ask what it’s borrowing. Is it a hedge that covers two audiences at once, a stack of negations, or a subject that shows up ten words late?
If you want a second pass beyond your own ear, I tested 13 AI detectors against known samples. Detectors measure this kind of structural rhythm, not single words.
The free copy cleaner checks a draft against these shapes plus the other fourteen signs in one pass.
What Does It Cost You to Miss These?
A reader isn’t grading your grammar. They’re deciding, inside a few seconds, whether a specific person wrote the page in front of them or a model filled in a template.
That decision happens before they’ve read your offer, your price, or a single proof point.
I watched a coaching client lose a warm lead over exactly this. Her homepage opened with a whether-or hedge and closed with a false range.
The prospect told her on the call that the page “felt like a template with her name on it.” She hadn’t said a word yet about her real method.
Nothing on that page was factually wrong. It just borrowed sentence shapes she’d never say out loud to a client sitting across from her.
A stranger caught it in the time it took to scroll once, and she never got the chance to correct the impression on the call.
I run these nine shapes, plus a broader 25-point editing pass, against every draft before it ships.
If a client’s draft has already cleared a word-list pass and still feels off, the shape underneath the words is usually why.
Fix the shape once, and it stays fixed no matter which words fill it in next time. That’s the real payoff of learning nine tropes instead of memorizing a longer and longer list of banned words.
FAQ
What does an AI sentence look like?
It stacks two or three of the same shape in a row, a whether-or hedge, a false range, or a negated pair, before it lands on the real point. The words can be completely different every time, and the shape still gives it away.
Are em dashes a dead giveaway for AI?
An em dash on its own just means someone owns a keyboard. The tell is density, three or four in one page welding half-formed ideas together, and I cover that pattern in depth in how to make AI writing sound more human.
How can you tell if a sentence was written by AI?
Check the shape before you check the words. A whether-or hedge, a stacked negation, or a gerund clause fronting the subject is a trope, not a coincidence, especially once it shows up twice on the same page.
What's the most overused AI writing trope?
Negative parallelism, stacking two negated claims before the real one, shows up on more landing pages than any other shape on this list, because it borrows the rhythm of a big realization while saying something ordinary.
What are some signs that AI wrote something?
The sentence shapes in this piece are one layer. The full list of fourteen signs of AI writing covers vocabulary, rhythm, and whole-page patterns too, since a single trope rarely travels alone.
How do you get rid of AI patterns in your writing?
Read the sentence out loud and name the shape it's borrowing. Once you can name the trope, cutting it takes one rewrite, not a rewrite of the whole paragraph.
What is parataxis, and why does it matter for AI writing?
Parataxis means placing clauses side by side with no word connecting them, so each fact sits alone instead of building on the last one. It drives several of the shapes on this list, and I go deeper on it in the sentence-rhythm breakdown.
What are some examples of AI writing tropes?
Negative parallelism, the false range, the inclusive whether-or hedge, and anaphora stacking are four of the nine I check for on every client draft, each with its own flagged example further down this page.
Want this built for your business? Book a free strategy call →