Skip to content

lguz/humanize-writing-skill / skills/humanize-writing

humanize-writing

Verified

Use when the user wants text to sound more human, says writing sounds "too AI" or "too ChatGPT," asks to humanize or rewrite a draft to feel natural, or shares content wanting it to feel authentic and less robotic. Also applies to LinkedIn posts, blog drafts, or emails where the user wants a more genuine voice.

FileKindSize
SKILL.mdskill md11.5 KB
references/ai-patterns-dictionary.mdmarkdown8.0 KB
references/voices.mdmarkdown6.7 KB

references/ai-patterns-dictionary.md

View on GitHub
# AI Writing Patterns Dictionary

A comprehensive reference of known AI writing tells, compiled from Wikipedia's "Signs of AI Writing," Reddit communities, GPTZero research, and professional editing forums. Use this as a checklist when reviewing and rewriting text.

---

## 1. Banned Vocabulary (High-Signal AI Words)

These words appear at statistically abnormal rates in LLM output vs. human writing. Avoid entirely or replace with simpler alternatives.

### Tier 1 — Strongest AI Signals (almost never used by humans in casual/professional writing)
| AI Word | Human Alternative |
|---------|-------------------|
| delve | dig into, look at, explore |
| tapestry | mix, combination |
| landscape | space, world, field |
| pivotal | important, key, big |
| underscore | show, highlight |
| testament | proof, sign, evidence |
| intricate / intricacies | complicated, detailed, tricky |
| meticulous / meticulously | careful, thorough |
| nuanced | subtle, complex |
| multifaceted | complex, many-sided |
| embark | start, begin |
| spearhead | lead, drive |
| bolster / bolstered | support, strengthen |
| garner | get, earn, attract |
| interplay | relationship, interaction |
| realm | area, space, world |
| labyrinth | maze, mess |
| symphony | mix, blend |

### Tier 2 — Moderate AI Signals (humans use them, but AI overuses them heavily)
| AI Word | Human Alternative |
|---------|-------------------|
| crucial | important, key |
| vibrant | lively, active |
| foster | encourage, support, grow |
| enhance | improve, boost |
| leverage | use, take advantage of |
| navigate | deal with, handle, figure out |
| resonate | connect, hit home, land |
| illuminate | clarify, show, explain |
| showcase | show, display |
| enduring | lasting, long-term |
| robust | strong, solid |
| holistic | complete, whole, full-picture |
| comprehensive | full, complete, thorough |
| innovative | new, creative, fresh |
| dynamic | active, changing, fluid |
| seamless / seamlessly | smooth, easy, without friction |
| cutting-edge | latest, newest, advanced |
| game-changer | big deal, breakthrough |

### Tier 3 — Transition Words (fine alone, but AI clusters them unnaturally)
| AI Transition | Human Alternative |
|---------------|-------------------|
| Furthermore | Also, And, Plus |
| Moreover | On top of that, And |
| Additionally | Also, And |
| Consequently | So, Because of this |
| Nevertheless | Still, But, Even so |
| Subsequently | Then, After that, Later |
| Notably | Worth mentioning |
| Indeed | (often just delete it) |
| Nonetheless | But, Still |
| Hence | So |
| Thus | So |
| In conclusion | (just conclude — don't announce it) |
| In summary | (just summarize — don't announce it) |
| It's worth noting that | (delete and just state the thing) |
| It's important to understand that | (delete and just state the thing) |

---

## 2. Banned Structural Patterns

These sentence/paragraph shapes are the strongest AI tells — even stronger than vocabulary.

### Parallel Negation ("Not X, but Y")
**AI pattern:** "Not because my standards dropped — but because I realized..."
**Why it's a tell:** LLMs love this rhetorical contrast structure. It appears 5-10x more often in AI text than human text.
**Fix:** Just state what happened. "I realized..." Skip the dramatic negation setup.

### Tricolon (Rule of Three)
**AI pattern:** "...research, collaboration, and problem-solving" / "vibrant, dynamic, and innovative"
**Why it's a tell:** LLMs default to groups of three adjectives, nouns, or phrases to sound comprehensive. Humans rarely write this way outside of speeches.
**Fix:** Pick the one or two that actually matter. Or just use a different sentence structure entirely.

### Em Dash Overuse
**AI pattern:** "The manager — who had been struggling — decided to act."
**Why it's a tell:** AI uses em dashes where humans would use commas, parentheses, or just split into two sentences. Wikipedia editors call this the "ChatGPT dash."
**Fix:** Use commas, periods, or parentheses. Reserve em dashes for rare emphasis only (max 1-2 per piece).

### The Rhetorical Question + Immediate Answer
**AI pattern:** "What does this mean? It means that..."
**Why it's a tell:** LLMs use this as a transition device constantly. Humans occasionally do it; AI does it every few paragraphs.
**Fix:** Just make the point directly.

### "Here's the thing:" / "Here's what nobody tells you:"
**AI pattern:** Starting paragraphs with dramatic reveals.
**Why it's a tell:** Overused LinkedIn/content-marketing AI pattern.
**Fix:** Drop the setup and start with the actual point.

### "Let's dive in" / "Let's unpack this" / "Let's break this down"
**AI pattern:** Invitation to explore.
**Why it's a tell:** Almost never appears in natural human writing. It's a podcast/YouTube intro structure.
**Fix:** Just start explaining.

### Neat Paragraph Endings
**AI pattern:** Every paragraph wraps up with a tidy conclusion or takeaway.
**Why it's a tell:** Humans leave loose ends, trail off, or let ideas hang. AI wraps everything in a bow.
**Fix:** Let some paragraphs just... end. Not every thought needs a landing.

### Mirror Structure (A/B Parallelism)
**AI pattern:** "A hands-on manager isn't micromanaging — they're firefighting. A distant manager isn't disengaged — they're trusting."
**Why it's a tell:** Perfect structural symmetry in consecutive sentences is an LLM favorite.
**Fix:** Break the symmetry. Make the second example structurally different from the first.

### Overuse of Colons for Reveals
**AI pattern:** "The result: a complete disaster." / "The outcome? Four managers agreed."
**Why it's a tell:** LLMs love the dramatic colon or question-mark reveal.
**Fix:** Weave it into a normal sentence.

### Inflation of Importance
**AI pattern:** "This represents a pivotal moment in..." / "This is a testament to..."
**Why it's a tell:** LLMs puff up the importance of everything. They rarely just state a fact without editorializing about how significant it is.
**Fix:** State facts. Let the reader decide if it's important.

---

## 3. Tone and Voice Tells

### Too Balanced / Too Fair
AI text gives equal weight to all perspectives. Real humans have opinions and lean into them.

### No Imperfections
Human writing has half-finished thoughts, slightly awkward transitions, sentences that start with "And" or "But," occasional fragments. AI text is uniformly polished.

### Absence of Personal Voice
AI text lacks the specific quirks (humor, sarcasm, frustration, tangents) that make writing feel human.

### Sycophantic / Over-Positive Tone
AI defaults to enthusiasm. "Great question!" "Absolutely!" "This is a fantastic approach!" Humans are more measured, skeptical, and sometimes blunt.

### Uniform Sentence Length
AI tends to write sentences of similar length throughout a piece. Humans naturally vary between short punchy sentences and longer meandering ones.

### Hedging Phrases
"Generally speaking," "to some extent," "from a broader perspective," "it could be argued that" — AI hedges constantly to appear balanced. Humans just say what they think.

---

## 4. Content-Level Red Flags

### Generic Conclusions
AI wraps up with vague, safe takeaways that could apply to almost anything: "Ultimately, what matters most is finding the right balance."

### No Concrete Details or Anecdotes
AI generates plausible-sounding but generic examples. Real human writing includes specific names, dates, places, numbers, and "you had to be there" details.

### Regression to the Mean
AI smooths specific facts into generic statements. Where a human might say "we lost $40k on that campaign," AI says "the initiative faced financial challenges."

---

## Sources

- Wikipedia: Signs of AI Writing (https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing)
- GPTZero AI Vocabulary Research (https://gptzero.me/ai-vocabulary)
- Chris Herbert's ChatGPT Overused Words (https://gist.github.com/chrisgherbert/c734ec50ae464135be57cd03b84281f9)
- Sabrina Ramonov Humanizer Prompt (https://github.com/SabrinaRamonov/prompts/blob/main/humanize_ai_writing.md)
- God of Prompt: 500 ChatGPT Overused Words (https://www.godofprompt.ai/blog/500-chatgpt-overused-words-heres-how-to-avoid-them)
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall