LLM cliché highlighter
Rhetorical tics
“No X, no Y” chains
Two or more “no …” items in a row, e.g. “No fluff, no filler, no jargon.” The badge counts the “no” items.
“That’s the whole …”
“That / this is the whole point, game, thing …”
“Did not X, did not Y” chains
Two or more “did not …” or “didn’t …” items in a row. The badge counts the items.
“Don’t VERB it … VERB it”
“Don’t call it X. Call it Y.” — a negated verb + “it”, then the same verb + “it” again.
“Sit with that”
The reflective “sit with that / this / it (for a moment)”, plus “sit with the discomfort” and friends.
“You already know”
“You already know” — the answer, what to do, or standing alone before a full stop.
“Is the entire …”
“X is the entire point / game / business model.”
“The entire … is”
“The entire point / game / business model is …” — the flipped twin of “is the entire”.
“Is real … and / not”
“The X is real, and / not …”, including “is the real … and it”. Skips “real estate”, “real time”, and similar.
“The punchline is”
“The punchline is …”, “the punchline:”, or “the punchline?”.
“Worth naming”
The therapist-voiced “that loss is real and it’s worth naming”, “it’s worth naming that …”, or a “Worth naming:” opener. Skips “naming names”.
“That’s not nothing”
“That is not nothing” / “that’s not nothing”, plus the “this / it / which is not nothing” variants.
“Is the whole …”
Any subject + “is the whole point / trick / pitch / idea”, plus the “here is the whole …” opener. The twin of “is the entire …”, and a generalisation of “That’s the whole …” to subjects other than that/this.
Echoing sentence runs
Consecutive sentences built on the same repeated skeleton — “A shopping cart is an object in the system. A chat room is an object in the system.” The badge counts the echoing sentences.
Performative honesty
Sincerity announced rather than demonstrated: “I won’t pretend”, “I’ll be honest”, “let’s be honest”, “to be clear”, and sentence-initial “Honestly,” or “Look,”.
“That’s the part …”
Gesturing at a favoured detail instead of stating it: “that is the part a counter can’t reach”, “the part that makes me trust the rest”, “my favourite part of …”.
“The only X I trust”
The narrowing superlative reveal: “the only marketing I trust”, “the only thing it needs”, “the only X that matters”.
“Don’t take my word for it”
The stock invitation to verify: “you don’t have to take my word for it”, “don’t take my word for any of this”.
“Turns out …”
The casual-revelation opener, almost always bolted to a tidy conclusion: “Turns out X”, “it turns out that X”.
“Fits in your head”
Dev-blog boilerplate for simplicity: “small enough to hold in your head”, “batteries included”, “it just works”, “zero config”, “sane defaults”.
Stacked rhetorical questions
Two or more questions fired in a row, usually fragments after the first: “Do I know how it works? Where it breaks? Which corners it cut?” The badge counts the questions.
Repeated sentence openers
Three or more consecutive sentences starting on the same word — “Maybe nobody needed it. Maybe it introduced … Maybe a small convenience …” Pronouns and articles are ignored. The badge counts the sentences.
Colon into a triple
A colon opening onto three or more comma-separated items: “separate ports, processes, and local state”. The most common shape LLM prose uses to sound concrete. Noisy in technical writing — leave it off by default if your corpus is documentation.
“Here’s the twist”
The stage-managed reveal: “here’s the twist”, “here’s the thing”, “here’s the catch / kicker / rub”, “here’s the first example:”.
“X is dead”
The obituary headline and its sequel: “peer code review is dead”, “botd is dead; long live botd”.
“That’s why X mattered”
Retroactively assigning significance: “that’s why being able to open the environment mattered”, “this is why preserving every conversation mattered”.
Stranded auxiliary contrast
A clause that lands on a bare auxiliary for the reversal: “The tool died; the data didn’t.”, “Reading mostly passed … Writing didn’t”, “Maybe it wouldn’t have.”
Signs of AI writing (Wikipedia)
AI vocabulary words
Words LLMs lean on far more than people do: “delve”, “tapestry”, “meticulous”, “pivotal”, “intricate”, “interplay”, “underscore”, “garner”, “bolster”, “vibrant”, “bustling”, “multifaceted”, “seamless”, “ever-evolving”. One hit can be coincidence — several is a tell.
“Not just X, but Y”
Negative parallelisms: “not just X, but (also) Y”, “not only … but …”, and the “it’s not X — it’s Y” contrast.
“It’s important to note”
Didactic hedging: “it is important to note that”, “it’s worth noting”, “it should be noted”, plus the “worth pausing / considering / asking” family.
“Stands as a testament”
“Stands / serves as a testament (or reminder)”, “is a testament to” — inflating significance instead of saying what happened.
“Plays a crucial role”
“Plays a crucial / pivotal / vital / key / significant role in …”.
“Ever-evolving landscape”
Scene-setting boilerplate: “the ever-evolving / changing / shifting landscape”, “in today’s fast-paced world”.
“Experts argue”
Vague attribution to unnamed authorities: “experts argue”, “some critics have noted”, “observers suggest”, “industry reports indicate”.
“Despite these challenges”
The boilerplate challenges-and-outlook formula: “despite these challenges”, “faces several challenges”, “challenges remain”, “remains to be seen”, “time will tell”.
Participle sentence tails
Superficial analysis bolted onto a sentence end: “…, highlighting / underscoring / showcasing / reflecting the …”.
Promotional boilerplate
Travel-brochure tone: “nestled in”, “in the heart of”, “rich tapestry / heritage”, “hidden gem”, “boasts a”, “breathtaking”, “stunning views”.
Chatbot leftovers
Artifacts pasted straight from a chatbot: “as an AI language model”, “as of my last update”, “knowledge cutoff”, plus markup debris like “oaicite”, “contentReference”, “turn0search” and “utm_source=” tracking parameters.
The second group is adapted from Wikipedia’s Signs of AI writing guide. More patterns can be added to the `patterns` array at the top of the script.