Signs of AI writing: 22 patterns and how to fix them
These are the habits that make text read as machine-written. Each guide shows what the pattern looks like, why readers notice it, before-and-after examples, and when it's fine to keep. Paste a draft into the checker to see which ones you have.
"It's not X, it's Y": The not-X-but-Y contrast is the most common tell in AI-written text. See examples, why it grates, and how to rewrite it in one pass.
"Delve" and the other AI words: Delve, tapestry, testament, seamless, leverage: the words that mark text as AI-generated, why models overuse them, and plain replacements.
Em dash overuse: Is using an em dash a sign of AI writing? Only when it's everywhere. How to spot dash overuse and what to use instead.
Forced triads (rule-of-three lists): "Innovation, inspiration, and insights." Why AI text arranges everything in threes, how to tell a real list from a forced one, and how to rewrite it.
Hedging openers and stacked qualifiers: "It's important to note that," "could potentially," "it's worth mentioning": the throat-clearing that makes AI text sound evasive, and how to cut it.
Dramatic one-line closers: "Let that sink in." "That's the real win." "Read that again." Why AI posts end paragraphs with a punchline that repeats the point, and how to fix it.
Rows of dramatic fragments: "No friction. No delays. No excuses." Why strings of three-word sentences read as AI-generated marketing, with fixes.
Faux-insight setups: "What nobody tells you is..." "The part everyone misses..." How AI text promises a secret and then delivers something ordinary.
Staged run-ups ("Let's dive in"): "Let's dive in." "Here's the thing." "In today's fast-paced world." Why AI intros announce the point instead of making it, and how to cut them.
Colon reveals: "The best part: it learns." "The catch:" "The result:" How staged reveals make copy sound generated, and simple rewrites.
Sayings that sound deep: "At its core," "the real question is," "X is the new oil": aphorisms that make AI text sound profound without saying much.
Arguing with no one: "Don't get me wrong," "to be clear," "you might think": defenses against objections nobody raised, and how to remove them.
Fake-profound endings: "The future is already here." "Exciting times ahead." "In a world where..." Why AI essays end on a grand line, and how to end on a fact instead.
Inflated significance: "Stands as a testament," "marks a pivotal moment," "plays a crucial role": how AI inflates ordinary facts, with plain rewrites.
Shallow -ing riders: ", highlighting," ", showcasing," ", underscoring": the participle tails AI adds to sound analytical, and how to cut them.
Sales language: "Nestled," "breathtaking," "world-class," "in the heart of": brochure words that make AI copy sound like an ad. Plain alternatives.
Borrowed authority: "Experts agree," "studies show," "industry reports suggest": vague sources that make AI text sound researched without being researched.
Avoiding "is" and "has": "Serves as," "stands as," "boasts": why AI replaces simple verbs with longer ones, and why plain verbs read better.
Bold-label bullet lists: "- **Speed:** Faster delivery." Why every-bullet-gets-a-bold-label formatting reads as ChatGPT output, and when to use prose instead.
Emoji bullets and arrows: 🚀 💡 👉 → Why lines that start with emojis or arrows read as AI-generated posts, and how to format without them.
Chatbot residue: "Great question!" "I hope this helps!" "Let me know if you'd like..." The chatbot leftovers people forget to delete, and how to catch them.
Knowledge-limit disclaimers: "As of my last training update," "while specific details are limited": model disclaimers that slip into published text, and what to write instead.