human-writing

name	human-writing
description	Use when writing prose, blog posts, articles, essays, or any non-code text content — removes AI patterns and produces natural human writing

Human Writing

Overview

AI-generated text has recognizable patterns that make readers instinctively distrust it. This skill identifies and eliminates those patterns so the output reads like a real person wrote it.

Core principle: Human writing is irregular, opinionated, and specific. AI writing is smooth, balanced, and generic. Kill the smoothness.

Banned Words and Phrases

These are the biggest AI tells. Never use them:

Dead Giveaway Words

Kill	Use instead
delve	dig into, look at, get into
landscape	(just drop it — "the security landscape" → "security")
robust	solid, strong, or just describe what makes it good
leverage	use
comprehensive	thorough, full, or drop it
crucial / vital	matters, important if you must
foster	build, grow, create
navigate	deal with, handle, figure out
underscores	shows, proves, means
multifaceted	complicated, messy
nuanced	(usually filler — cut the whole sentence)
pivotal	important, key
commendable	(don't praise — just state what happened)
noteworthy	(same — state the fact, let the reader decide)
intricate	complicated, messy, detailed
realm	area, world, space
tapestry	(never)
embark	start, begin
utilizing	using
encompass	cover, include
seamless	smooth, easy, or describe what makes it smooth
streamline	simplify, speed up
harness	use, take advantage of
spearhead	lead, start
bolster	strengthen, support
elevate	improve, raise
amplify	increase, spread
groundbreaking	new, first, or describe what's actually new
cutting-edge	new, latest
game-changing	(describe the actual change instead)
transformative	(same — show the transformation, don't label it)
revolutionary	(same)
meticulous	careful, thorough
aforementioned	(just name the thing again)
pertinent	relevant
holistic	(usually meaningless — cut it)
synergy	(never)
resonate	connect, stick, land
align	match, fit
empower	let, enable, give
unlock	enable, open up
supercharge	speed up, improve

Filler Phrases to Delete

• "It's important to note that" → just state the thing
• "It's worth mentioning that" → just mention it
• "In today's rapidly evolving..." → delete the whole sentence
• "At the end of the day" → cut it
• "When it comes to..." → just start with the topic
• "In order to" → "to"
• "It goes without saying" → then don't say it
• "Let's dive in" / "Let's delve into" → just start
• "Without further ado" → just start
• "In the realm of" → "in"
• "A myriad of" → "many" or "lots of"
• "First and foremost" → "First" or just start
• "Last but not least" → "Finally" or just state it
• "Moving forward" → cut it
• "That being said" → "But" or cut it
• "At its core" → cut it or just say the thing
• "Plays a crucial role in" → "matters for" or restructure
• "Serves as a testament to" → "shows", "proves"
• "Paves the way for" → "leads to", "enables"
• "Sheds light on" → "shows", "explains", "clarifies"
• "A wide range of" → "many", "lots of"
• "On the other hand" (every single time there's a counterpoint) → "But", or just state the counterpoint
• "Deep dive" → "look", "analysis", or just do the analysis
• "Double-edged sword" → describe the tradeoff directly
• "At the heart of" → cut it, just name the thing

Don't Use Emoji in Prose

Emoji in articles, blog posts, and essays is an instant AI tell. Real writers don't sprinkle 🚀🔥💡✨ through their paragraphs. If you're writing a serious piece, zero emoji. If you're writing something casual (like a tweet or chat message), one or two max — and only if it adds meaning, not decoration.

Specific bans:

• Never use emoji as bullet points (❌ ✅ 🔹)
• Never use emoji to "emphasize" a heading or section title
• Never use 🚀 to mean "launch/ship" or 💡 to mean "idea" or 🔥 to mean "great"
• Never stack multiple emoji together (🎉🚀✨)
• The clapping-between-words pattern (👏 Don't 👏 Do 👏 This) is also banned

Banned Transitions (overused by AI)

• "Moreover" → "And" or restructure
• "Furthermore" → "And" or just start the next point
• "Additionally" → "Also" or just state it
• "Consequently" → "So"
• "Nevertheless" → "But", "Still"
• "Subsequently" → "Then", "After that"
• "Henceforth" → (never)
• "Notwithstanding" → "Despite", "But"

Tone Tells to Avoid

• Relentlessly positive — never says anything is bad. Sometimes things are bad. Say so.
• Perfectly balanced on everything — "While X has merit, Y also has advantages." Pick a side.
• Excessive hedging — "somewhat", "potentially", "arguably", "it could be said that". Delete half your qualifiers.
• Passive voice to dodge opinions — "it can be observed that" → "I noticed"
• Uncontracted in casual prose — "do not", "it is", "they are" → "don't", "it's", "they're"
• Over-polite/diplomatic — won't commit to a position on anything
• Generic praise — "excellent", "remarkable", "impressive" without saying what's actually good about it

Structural Anti-Patterns

Document-Level

The AI template — almost all AI writing follows this skeleton:

• "What is X?" → "Why X matters" → "How to use X" → "Conclusion"
• Break this pattern. Start with a story. Start with the punchline. Start with the controversial take.

Always exactly 3 main points — AI loves the number 3. Sometimes the answer is 2 things. Sometimes it's 7. Don't force everything into trinities.

Intro announces structure — "In this article, we will explore three key aspects of..." Just start talking about the thing.

Conclusion restates everything — AI always ends by summarizing what it just said. Humans end when they're done. The last point IS the ending. Or end with a concrete action. Or end mid-thought.

"In conclusion" / "To summarize" / "In summary" as final heading — If you need to label your conclusion, the essay failed. Just end it.

Every section roughly the same length — If your 4 sections are all ~300 words each, something is wrong. Some ideas need more space. Some need less.

Perfectly symmetrical organization — If section 2 has 3 sub-points, AI makes sections 1 and 3 also have 3 sub-points. Real writing is asymmetric.

Paragraph-Level

Uniform paragraph length — all ~3-5 sentences. Humans write paragraphs of wildly different lengths.

Some paragraphs are one sentence.

Others go on for a while because the idea is complex and needs room to breathe, and maybe there's a tangent in there too because that's how thinking actually works — you don't always organize your thoughts into perfectly bite-sized chunks before you express them.

Topic sentence → support → transition, every time — This is the five-paragraph-essay formula. It works for school. It sounds robotic in real writing.

Every paragraph connects smoothly to the next — Sometimes you just... change the subject. That's fine. Readers can handle it.

Sentence-Level

Similar sentence lengths throughout — no rhythm variation. Real writing alternates. Short ones hit hard. Then longer ones build and develop and add nuance.

Excessive parallelism in lists — "X enables Y. Z enables W. A enables B." Same structure repeated makes it sound generated.

Starting consecutive sentences the same way — "This means... This allows... This ensures..." Vary your openings.

List-Level

Always bullet points, never inline — Sometimes a list works better as a sentence: "You'll need a hammer, some nails, and questionable judgment."

Every bullet same length — Vary it. Some bullets are one word. Some are a sentence.

Exactly 3 or 5 items — Real lists have however many items the topic requires.

Bullets that each start with a bold keyword followed by colon and explanation — "Security: We ensure..." / "Performance: We optimize..." / "Scalability: We design..." This is the most AI-looking list format possible.

Opening Tells

• Starting with a definition: "X is a Y that Z" → don't. Start with a story, a question, or a claim.
• "Imagine a world where..." → just describe the thing
• "Have you ever wondered..." → ask the real question or just state it
• "In the ever-changing world of..." → delete
• Rhetorical question + "The answer might surprise you" → clickbait AI pattern

Closing Tells

• Restating all points made above
• "The future of X is bright" → (never)
• Call to action that sounds like marketing copy
• "Only time will tell" → (never)
• Ending on an inspirational note regardless of topic
• "By following these steps, you can..." → just end after the last step

Meta-Commentary to Cut

• "This is a great question."
• "There are several factors to consider." → just discuss the factors
• "Let me explain why this matters." → just explain
• "Let's take a closer look at..." → just look at it
• "Before we begin..." / "Prerequisites" / "What you'll learn" sections in essays

Voice and Tone Rules

Be specific, not generic

• Bad: "This can have significant implications for organizations."
• Good: "This means your deploy pipeline breaks at 3am on a Friday."

Use concrete examples from real life

AI invents generic scenarios. Humans reference specific things — products, companies, experiences, cultural touchstones. Say "like the time Cloudflare took down half the internet" not "like when a major service provider experiences an outage."

Contractions are mandatory

"It's", "don't", "won't", "can't", "they're". Not "it is", "do not", "will not." Uncontracted prose in informal writing is an instant AI tell.

Vary sentence length aggressively

AI writes sentences of similar length. Humans don't.

Short ones hit hard.

Then sometimes you write a longer one that builds on the short one and adds context and maybe goes on a bit and that's fine because rhythm in writing comes from contrast, not from consistency.

Then short again.

Use "I" and have opinions

AI hides behind passive voice and "we." Pick a side. "I think this is wrong" is better than "Some might argue there are potential concerns."

Don't be relentlessly positive

Sometimes things are bad. Say so. "This API is terrible" is more honest than "This API presents some challenges but also offers opportunities for creative solutions."

Self-Check Before Publishing

Read it out loud. If you wouldn't say it in a conversation, rewrite it. Specifically:

1. Search for banned words — ctrl+F for "delve", "landscape", "robust", "leverage", "seamless", "holistic", "synergy", etc.
2. Count your transitions — if you have more than one "Moreover/Furthermore/Additionally" per 500 words, rewrite them.
3. Check your ending — if the last paragraph restates everything above, delete it.
4. Check paragraph lengths — if more than 3 consecutive paragraphs are similar length, vary them.
5. Check for hedging — if every claim has a qualifier ("somewhat", "potentially", "arguably"), delete half of them.
6. Check for the "three things" pattern — if you have exactly 3 of everything, change some of them.
7. Check list formatting — if every bullet starts with Bold Keyword: followed by explanation, restructure.
8. Check openings — if you started with a definition or "Imagine...", rewrite the first paragraph.
9. Check for emoji — if any emoji appear in prose, remove them all.

Common Mistakes

Mistake	Fix
Overcorrecting into try-hard casualness ("Yo, let's talk about databases!")	Write like a smart person talking to another smart person. Casual ≠ juvenile.
Removing ALL structure	Structure is fine. Just don't use the same structure every time.
Forcing jokes	If a joke comes naturally, keep it. Don't insert humor on a schedule.
Replacing AI cliches with human cliches	"At the end of the day" is just as bad coming from a human. Say something specific instead.
Making everything angry/edgy to sound "authentic"	Opinions ≠ anger. You can disagree calmly.