humanizer
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Calibrates output to document type (CV, LinkedIn post, blog, email,
SKILL.mdskills/humanizer/SKILL.md
GitHubHumanizer: remove AI writing patterns
Rewrite AI-sounding text so it reads like the writer, not a chatbot. Do not change what it says or make up details.
STEP 0 - Calibrate to Document Type (DO THIS FIRST)
Before applying any pattern below, identify what kind of text you're editing. Different formats have different rules. Applying blog-post rhythm to a CV destroys the CV. Applying CV concision to a blog post makes it lifeless.
Identify the document type, then load the matching profile:
CV / Resume bullets
- Sentence length: Mixed - let the content decide. Some bullets are tight one-liners (12-20 words: action + result). Others need context and naturally run 20-35 words. Do not artificially shorten a bullet that needs context.
- Structure: Each bullet is one connected unit. Do NOT break a bullet into two short sentences. A bullet that reads "Led X campaign - 12B VND budget - top-3 ranking" should become "Led the 12B VND X campaign that ranked top-3 nationally for three consecutive years", NOT "Led X campaign. Budget: 12B VND. Ranked top-3."
- Voice: Active verbs, professional, no first-person "I" needed. Past tense for past roles, present for current.
- Parallelism: Bullets within a section should have similar grammatical shape. This is intentional, not "AI-sounding." Do not break parallelism just because the patterns below say so.
- What to skip from the patterns below: "Vary your rhythm" (parallelism wins here). "Add personality / first-person." "Let some mess in." Most "soul" advice does not apply.
- What to keep: All AI-vocabulary checks (testament, pivotal, landscape, leverage, etc.), promotional language, copula avoidance, em/en dash overuse, hyphenated word pairs, vague sources, generic conclusions.
Cover letter / Personal statement
- Sentence length: Mostly 18-30 words. A short sentence for emphasis is fine, but not three in a row.
- Voice: First person, warm-professional, specific. The opening should NOT be "I am writing to apply for..." but it also should not be a one-line punchline.
- What to keep from "soul" guidance: Specificity, concrete examples. Skip overt opinions and tangents.
LinkedIn post / Professional social
- Sentence length: Varied. Mix of medium (15-25 words) and short (5-10) for emphasis. Avoid three short sentences in a row unless building to something.
- Voice: First person, conversational-professional. Some opinion is fine.
- Apply most patterns below. This is the format the patterns are most calibrated for.
Blog post / Article / Essay
- Sentence length: Genuinely varied. Long flowing sentences (30+ words) are fine and often necessary. Short punchy ones for impact. The rhythm should feel like a person thinking, not a metronome.
- Voice: Full personality. First person, opinions, asides, even mild tangents.
- Apply all patterns below, including the "soul" section.
Email / Slack message
- Sentence length: Match the recipient's register. Internal Slack: short. Client email: medium, complete.
- Voice: Direct. No throat-clearing ("I hope this message finds you well"). No sign-off theater.
- Apply the chatbot-artifact and disclaimer patterns hardest.
Press release / Formal report
- Sentence length: Medium-to-long (20-35 words). Complete clauses. No fragments.
- Voice: Third person, neutral, factual. No personality injection.
- What to skip: All "soul" guidance. First-person voice. Opinion injection.
- What to keep: Promotional language removal, vague sources, generic conclusions, AI vocabulary.
Marketing copy / Ad
- Sentence length: Whatever sells. Fragments are intentional. Short punchy lines are intentional. Do NOT "fix" these.
- What to skip: Sentence-length rules. Passive-voice flagging if active is too clunky.
- What to keep: Genuine AI tells (testament, vibrant, nestled, etc.) that read as fake-corporate even in ad copy.
If unsure
Ask the user once: "What format is this - CV bullet, LinkedIn post, blog, email, or something else?" Don't guess.
CRITICAL: The "Vary Your Rhythm" Anti-Pattern
The single biggest failure mode of this skill historically has been over-fragmenting content into too many short sentences. This is itself an AI tell - content that reads like a list rather than connected prose.
Rules to prevent this:
- Never produce three short sentences (under 12 words each) in a row unless it's marketing copy or genuinely punchy emphasis.
- A single connected idea stays in one sentence. Do not split "X happened because Y, which led to Z" into three sentences just to "vary rhythm."
- Subordinate clauses are your friend. "While leading the campaign, the team grew from 4 to 12 and revenue tripled" beats "The campaign was led. The team grew from 4 to 12. Revenue tripled."
- Read your output aloud mentally. If it sounds like Morse code (dot dot dot dot), it's wrong. If it sounds like a person speaking, it's right.
- Long sentences are not AI-sounding by themselves. A 40-word sentence with real subordinate clauses, specific detail, and a clear arc is a human sentence. AI tells are about vocabulary, vagueness, and empty significance - not length.
See pattern 36 (Over-fragmentation) below for the full write-up.
What to do
When given text to humanize:
- Calibrate to document type (Step 0 above).
- Find AI patterns. Check the text against the patterns below, applying only those relevant to the format.
- Keep every claim. You may shorten dull parts, expand useful parts, and merge or split paragraphs. Keep the information even when you change the structure.
- Do not invent facts. Do not add a fact, name, number, date, quote, or citation unless it comes from the source or the user. If a sentence needs a missing detail, ask for it or use a simpler sentence. You may add an opinion or reaction when the writer's voice calls for one, but you may not add a factual claim. Fiction is exempt because invented details are part of the task.
- Match the voice. Use the right tone for the text, such as formal, casual, or technical. Add personality only when the text, the writer, and the document-type profile call for it.
- Preserve format conventions. A CV stays a CV; do not blog-post-ify it.
- Check for over-fragmentation. Count consecutive short sentences. Merge if you find three in a row.
- Do a final anti-AI pass. Ask "What still reads as AI here?" Answer briefly, then revise.
The input type controls what you return. See How to return the result. Use the same rewrite process in every mode.
Match the writer's voice
If the user provides a writing sample (their own previous writing), analyze it before rewriting:
- Read the sample first. Note its sentence length, word choice, paragraph openings, punctuation habits (dashes? parenthetical asides? semicolons?), recurring phrases or verbal tics, and transitions (explicit connectors, or just starting the next point?).
- Match those habits. Do not replace casual words with formal ones or remove deliberate quirks. If they write long flowing sentences, don't chop them into short ones. If they use "stuff" and "things," don't upgrade to "elements" and "components."
- If there is no sample, fall back to the document-type profile from Step 0.
How to provide a sample:
- Inline: "Humanize this. Here's a sample of my writing: [sample]"
- File: "Humanize this. Use my writing style from [file path] as a reference."
A writing sample takes priority over these style rules. If the sample uses em dashes, keep them at about the same rate. Do not apply pattern 14 as a ban.
Personality and soul (apply selectively - see Step 0)
This section applies to blog posts, LinkedIn posts, personal essays, and other voice-driven formats. Skip for CVs, press releases, formal reports, and most professional documents.
Removing AI patterns is only half the job. The result should still sound like a person - sterile, voiceless writing is just as obvious as slop.
Signs of soulless writing (in voice-driven formats):
- Every sentence is the same length and structure
- No opinions, just neutral reporting
- No acknowledgment of uncertainty or mixed feelings
- No first-person perspective when appropriate
- No humor, no edge, no personality
- Reads like a Wikipedia article or press release
How to add voice (where appropriate):
Have opinions. Don't just report facts - react to them. "I genuinely don't know how to feel about this" is more human than neutrally listing pros and cons.
Vary your rhythm - but don't over-fragment. Mix sentence lengths. The mix is medium-with-occasional-short, not short-short-short. See the anti-pattern callout above.
Use "I" when it fits. First person isn't unprofessional - it's honest. Skip this for CVs and formal reports.
Let some mess in. Perfect structure feels algorithmic. Tangents, asides, and half-formed thoughts are human - in blogs and posts, not in CVs or reports.
Use personality in blog posts, essays, opinions, and personal writing when it fits the writer. Keep reference, technical, legal, and factual text neutral. Do not add opinions or first-person language where they do not belong. When personality fits, keep the writer's opinions, uncertainty, mixed feelings, humor, asides, and uneven rhythm. Never invent facts to make the text feel personal.
Before (clean but soulless - blog post):
The experiment produced interesting results. The agents generated 3 million lines of code. Some developers were impressed while others were skeptical. The implications remain unclear.
After (has a pulse - blog post):
I genuinely don't know how to feel about this one. 3 million lines of code, generated while the humans presumably slept, and half the dev community is losing their minds while the other half is busy explaining why it doesn't count. The truth is probably somewhere boring in the middle, but I keep thinking about those agents working through the night.
Content patterns
1. Inflated claims about importance and legacy
Words to watch: stands/serves as, is a testament/reminder, a vital/significant/crucial/pivotal/key role/moment, underscores/highlights its importance/significance, reflects broader, symbolizing its ongoing/enduring/lasting, contributing to the, setting the stage for, marking/shaping the, represents/marks a shift, key turning point, evolving landscape, focal point, indelible mark, deeply rooted Problem: AI writing often claims that ordinary details mark a major change, prove a legacy, or reflect a broad trend. Before:
The Statistical Institute of Catalonia was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain. This initiative was part of a broader movement across Spain to decentralize administrative functions and enhance regional governance. After: The Statistical Institute of Catalonia was established in 1989, part of a wider decentralization of administrative functions in Spain.
2. Name-dropping to prove importance
Words to watch: independent coverage, local/regional/national media outlets, written by a leading expert, active social media presence Problem: AI writing often lists well-known publications or follower counts to prove that a person matters. The list usually gives no useful context. Before:
Her views have been cited in The New York Times, BBC, Financial Times, and The Hindu. She maintains an active social media presence with over 500,000 followers. After: Her views have been cited in The New York Times and the BBC.
If the source explains what the person said and where, keep that useful citation. Do not invent context for a shorter version.
3. Shallow analysis with -ing phrases
Words to watch: highlighting/underscoring/emphasizing..., ensuring..., reflecting/symbolizing..., contributing to..., cultivating/fostering..., encompassing..., showcasing... Problem: AI writing often adds an -ing phrase to make a simple fact sound deeper than it is. Before:
The temple's color palette of blue, green, and gold resonates with the region's natural beauty, symbolizing Texas bluebonnets, the Gulf of Mexico, and the diverse Texan landscapes, reflecting the community's deep connection to the land. After: The temple is painted blue, green, and gold, colors meant to evoke Texas bluebonnets and the Gulf of Mexico.
Note: Do not turn this into "The temple uses three colors. Blue. Green. Gold. They reference bluebonnets." That is over-fragmentation (pattern 36), which is its own AI tell.
4. Promotional and Advertisement-like Language
Words to watch: boasts a, vibrant, rich (figurative), profound, enhancing its, showcasing, exemplifies, commitment to, natural beauty, nestled, in the heart of, groundbreaking (figurative), renowned, breathtaking, must-visit, stunning Problem: AI writing often sounds like an advertisement, especially when it describes places, culture, products, or organizations. Before:
Nestled within the breathtaking region of Gonder in Ethiopia, Alamata Raya Kobo stands as a vibrant town with a rich cultural heritage and stunning natural beauty. After: Alamata Raya Kobo is a town in the Gonder region of Ethiopia.
5. Vague sources
Words to watch: Industry reports, Observers have cited, Experts argue, Some critics argue, several sources/publications (when few cited) Problem: AI writing often assigns a claim to unnamed experts, critics, reports, or observers. Before:
Due to its unique characteristics, the Haolai River is of interest to researchers and conservationists. Experts believe it plays a crucial role in the regional ecosystem. After: Researchers and conservationists study the Haolai River for its unusual characteristics.
Name a real source when the source text provides one. Otherwise, remove the unsupported claim. Never invent a source.
6. Formulaic challenges and outlook sections
Words to watch: Despite its... faces several challenges..., Despite these challenges, Challenges and Legacy, Future Outlook Problem: AI articles often add a stock section about challenges, future prospects, or continued growth. These sections usually repeat vague claims instead of adding facts. Before:
Despite its industrial prosperity, Korattur faces challenges typical of urban areas, including traffic congestion and water scarcity. Despite these challenges, with its strategic location and ongoing initiatives, Korattur continues to thrive as an integral part of Chennai's growth. After: Korattur has recurring traffic congestion and water shortages.
Add details such as dates or public actions only when they come from the source or the user.
Language and grammar patterns
7. Overused AI words
High-frequency AI words: Actually, additionally, align with, crucial, delve, emphasizing, enduring, enhance, fostering, garner, gate/gated/gating (figurative; preserve established technical usage), highlight (verb), interplay, intricate/intricacies, key (adjective), landscape (abstract noun), leverage (verb), pivotal, quietly, robust, seamless, showcase, tapestry (abstract noun), testament, underscore (verb), valuable, vibrant Problem: AI writing uses these words much more often than most people do, especially in groups. CV note: "leverage," "robust," "seamless," "key" (adjective), and "valuable" are especially overused on CVs. Replace with concrete verbs and specifics. Before:
Additionally, a distinctive feature of Somali cuisine is the incorporation of camel meat. An enduring testament to Italian colonial influence is the widespread adoption of pasta in the local culinary landscape, showcasing how these dishes have integrated into the traditional diet. After: Somali cuisine also includes camel meat, which is considered a delicacy. Pasta dishes, introduced during Italian colonization, remain common, especially in the south.
8. Avoiding is and are
Words to watch: serves as/stands as/marks/represents [a], boasts/features/offers [a] Problem: AI writing often replaces simple verbs such as is, are, and has with longer phrases. Before:
Gallery 825 serves as LAAA's exhibition space for contemporary art. The gallery features four separate spaces and boasts over 3,000 square feet. After: Gallery 825 is LAAA's exhibition space for contemporary art. The gallery has four rooms totaling 3,000 square feet.
9. Not X but Y and clipped negative endings
Problem: AI writing overuses forms such as "Not only...but..." and "It's not just X, it's Y." It also adds clipped endings such as "no guessing" or "no wasted motion" instead of writing a clear clause. Before:
It's not just about the beat riding under the vocals; it's part of the aggression and atmosphere. It's not merely a song, it's a statement. After: The heavy beat adds to the aggressive tone. Before (tailing negation): The options come from the selected item, no guessing. After: The options come from the selected item without forcing the user to guess.
10. Forced groups of three
Problem: AI writing often forces ideas into groups of three to sound complete. Before:
The event features keynote sessions, panel discussions, and networking opportunities. Attendees can expect innovation, inspiration, and industry insights. After: The event includes talks and panels. There's also time for informal networking between sessions.
A real triple from real life is fine. The problem is forced triples where the third item is filler.
11. Changing names and repeating sentence openings
Problem: AI writing handles repetition by rule instead of by ear. It may keep renaming the same person or thing. It may also start several sentences with the same subject, often she or he.
Use one clear name for the same subject. For repeated openings, merge sentences, change the subject when that helps, or begin with the action. Before (synonym cycling):
The protagonist faces many challenges. The main character must overcome obstacles. The central figure eventually triumphs. The hero returns home. After: The protagonist faces many challenges but eventually triumphs and returns home. Before (repeated openings): She noted the door. She noted the lock on it. She filed both away. After: She noted the door and its lock, then filed both away.
Do not ban the repeated word. Fix the repeated sentence pattern. The remaining sentence may still start with "She."
12. False from X to Y ranges
Problem: AI writing often uses "from X to Y" when X and Y do not form a real range. Before:
Our journey through the universe has taken us from the singularity of the Big Bang to the grand cosmic web, from the birth and death of stars to the enigmatic dance of dark matter. After: The book covers the Big Bang, star formation, and current theories about dark matter.
13. Passive voice and missing subjects
Problem: AI writing often hides who acts or drops the subject. Use active voice when it makes the actor and action clearer. Format note: CVs use passive sparingly - active verbs lead each bullet. Press releases and formal reports use passive when the actor genuinely doesn't matter. Before:
No configuration file needed. The results are preserved automatically. After: You do not need a configuration file. The system preserves the results automatically.
Style patterns
14. Em and en dashes
Rule: The final rewrite must not contain em dashes (—) or en dashes (–), unless the writer's sample uses them. Replace a dash with a period, comma, colon, or parentheses, or rewrite the sentence. Also check for spaced dashes (—) and double hyphens (--) used as dashes.
Before:
The term is primarily promoted by Dutch institutions—not by the people themselves. You don't say "Netherlands, Europe" as an address—yet this mislabeling continues—even in official documents. After: The term is primarily promoted by Dutch institutions, not by the people themselves. You don't say "Netherlands, Europe" as an address, yet this mislabeling continues in official documents. Before: The new policy — announced without warning — affects thousands of workers. The changes -- long overdue according to critics -- will take effect immediately. After: The new policy, announced without warning, affects thousands of workers. The changes, long overdue according to critics, will take effect immediately.
Before returning the rewrite, search for — and –. Remove each one unless the writer's sample uses that mark. In that case, match the sample's rate.
15. Too much bold text
Problem: AI chatbots often bold words and phrases without a clear reason. Before:
It blends OKRs (Objectives and Key Results), KPIs (Key Performance Indicators), and visual strategy tools such as the Business Model Canvas (BMC) and Balanced Scorecard (BSC). After: It blends OKRs, KPIs, and visual strategy tools like the Business Model Canvas and Balanced Scorecard.
16. Lists with bold mini-headings
Problem: AI writing often uses vertical lists in which every item starts with a bold label and a colon. Format caveat: CVs legitimately use bullet lists. Do NOT flatten CV bullets into prose. The pattern below applies to blog posts, articles, and emails where AI inserted a fake list to look organized. Before:
- User Experience: The user experience has been significantly improved with a new interface.
- Performance: Performance has been enhanced through optimized algorithms.
- Security: Security has been strengthened with end-to-end encryption. After: The update improves the interface, speeds up load times through optimized algorithms, and adds end-to-end encryption.
17. Title case in headings
Problem: AI chatbots often capitalize every main word in a heading. Before:
Strategic Negotiations And Global Partnerships
After:
Strategic negotiations and global partnerships
18. Emojis
Problem: AI chatbots often add emojis to headings and list items as decoration. Before:
🚀 Launch Phase: The product launches in Q3 💡 Key Insight: Users prefer simplicity ✅ Next Steps: Schedule follow-up meeting After: The product launches in Q3. User research showed a preference for simplicity. Next step: schedule a follow-up meeting.
19. Curly quotation marks
Problem: ChatGPT often uses curly quotes (“...”) where the writer or target format uses straight quotes ("..."). Before:
He said “the project is on track” but others disagreed. After: He said "the project is on track" but others disagreed.
Chatbot patterns
20. Chatbot text left in the answer
Words to watch: I hope this helps, Of course!, Certainly!, You're absolutely right!, Would you like..., Want me to...?, Want me to give examples?, Should I continue?, let me know, here is a... Problem: A chatbot's greeting, offer, or closing sometimes remains in text that should stand on its own. Before:
Here is an overview of the French Revolution. I hope this helps! Let me know if you'd like me to expand on any section. After: The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest.
21. Knowledge-limit disclaimers and guesses
Words to watch: as of [date], Up to my last training update, While specific details are limited/scarce..., based on available information, not publicly available, maintains a low profile, keeps personal details private, prefers to stay out of the spotlight, likely [grew up/studied/began], it is believed that Problem: Older models may mention the date when their knowledge ends. A model may also explain that it could not find a source, then fill the gap with a plausible guess. State what the source does not show, or remove the sentence. Do not present a guess as a fact. Before (cutoff disclaimer):
While specific details about the company's founding are not extensively documented in readily available sources, it appears to have been established sometime in the 1990s. After: The company's founding date is not documented in the available sources. (Or cut the sentence. State a date only if a source provides one.) Before (speculative gap-fill): Information about her early life is not publicly available, suggesting she maintains a low profile and keeps personal details private. She likely grew up in a middle-class household, which shaped her later interest in education reform. After: Her early life is not documented in the available sources. (Or omit the section.)
22. Overly agreeable tone
Problem: AI assistants often praise the user or agree before giving the answer. Before:
Great question! You're absolutely right that this is a complex topic. That's an excellent point about the economic factors. After: The economic factors you mentioned are relevant here.
Filler and hedging
23. Filler phrases
Before → After:
- "In order to achieve this goal" → "To achieve this"
- "Due to the fact that it was raining" → "Because it was raining"
- "At this point in time" → "Now"
- "In the event that you need help" → "If you need help"
- "The system has the ability to process" → "The system can process"
- "It is important to note that the data shows" → "The data shows"
24. Too many qualifiers
Phrases to watch: to be fair, it's also possible, could potentially, might arguably, in some cases it may, this is an inference Problem: Repeated editing can add one qualifier after another until every claim sounds uncertain. Keep a qualifier only when the source supports it and the meaning needs it. Remove caveats that only repair an earlier overstatement. Before:
It could potentially possibly be argued that the policy might have some effect on outcomes. After: The policy may affect outcomes.
25. Generic positive endings
Problem: AI writing often ends with vague optimism instead of the last useful fact. Before:
The future looks bright for the company. Exciting times lie ahead as they continue their journey toward excellence. This represents a major step in the right direction. After: (Cut the paragraph. End on the last concrete fact instead of a send-off. If the source states real plans, use those.)
26. Too many hyphenated word pairs
Words to watch: third-party, cross-functional, client-facing, data-driven, decision-making, well-known, high-quality, real-time, long-term, end-to-end
Problem: AI writing often hyphenates these pairs everywhere. Keep the hyphen before a noun when grammar needs it, as in a high-quality report. Drop it after the noun, as in the report is high quality.
CV note: A few hyphenated compounds in a CV are fine and expected. The flag is when every compound modifier in the document is hyphenated - that's the AI tell.
Before:
The cross-functional team delivered a high-quality, data-driven report. The team is cross-functional, the report is high-quality, and the methodology is data-driven. After: The cross-functional team delivered a high-quality, data-driven report. The team is cross functional, the report is high quality, and the methodology is data driven.
27. Persuasive authority tropes (pretending to reveal a deeper truth)
Phrases to watch: The real question is, at its core, in reality, what really matters, fundamentally, the deeper issue, the heart of the matter Problem: AI writing uses these phrases to make an ordinary point sound like a hidden truth. Before:
The real question is whether teams can adapt. At its core, what really matters is organizational readiness. After: The question is whether teams can adapt. That mostly depends on whether the organization is ready to change its habits.
28. Announcing the next point
Phrases to watch: Let's dive in, let's explore, let's break this down, here's what you need to know, now let's look at, without further ado, heads up, quick note, before I forget Problem: AI writing often announces the next point instead of stating it. A casual phrase such as "one thing that bit me" can have the same problem. Remove the announcement, not just its formal tone. Before:
Let's dive into how caching works in Next.js. Here's what you need to know. After: Next.js caches data at multiple layers, including request memoization, the data cache, and the router cache. Before (casual register): One thing that bit me hard, so pay attention to this part: the webpack dev server doesn't send the CORS header by default. After: The webpack dev server doesn't send the CORS header by default.
29. A heading repeated in the first sentence
Signs to watch: A heading followed by a one-line paragraph that simply restates the heading before the real content begins. Problem: AI writing often follows a heading with a sentence that only repeats the heading. Remove the repeated sentence. Before:
Performance
Speed matters.
When users hit a slow page, they leave. After:
Performance
When users hit a slow page, they leave.
30. Writing about the previous version
Problem: Documentation and comments should describe the current behavior. Mention the previous version only in change logs, release notes, migration guides, and other documents about change. Before:
This function was added to replace the previous approach of iterating through all items, which caused O(n²) performance. After: This function uses a hash map for O(1) lookups, avoiding the O(n²) cost of naive iteration.
31. Forced punchlines and dramatic fragments
Problem: AI writing often turns each sentence into a dramatic closing line. One short sentence can add emphasis. A row of short fragments usually feels forced. Before:
Then AlphaEvolve arrived. It had no preference for symmetry. No aesthetic prior. No nostalgia for human taste. The old rules were gone. After: AlphaEvolve changed the search because it did not favor symmetry or human-looking designs. That made some of the older assumptions less useful.
32. Formulaic sayings
Words to watch: X is the Y of Z, X becomes a trap, X is not a tool but a mirror, the language of, the currency of, the architecture of Problem: AI writing often turns an ordinary claim into a saying that sounds deep but adds no detail. Replace the saying with the specific claim. Before:
Symmetry is the language of trust. Efficiency becomes a trap when teams forget the human layer. After: Symmetric layouts often feel more predictable to users. Teams can over-optimize workflows and miss how people actually use them.
33. Fake-candid openings
Phrases to watch: Honestly?, Look, Here's the thing, The thing is, Let's be honest, Real talk, when used as standalone hooks or fake-candid pauses before an ordinary point. Problem: AI writing often starts with a staged pause or claim of honesty before making a routine point. State the point directly. Before:
Is it worth the price? Honestly? It depends on how often you'll use it. After: Whether it's worth the price depends on how often you'll use it.
34. Answering objections no one raised
Phrases to watch: This isn't (mainly/really) about, I'm not saying/arguing/trying to, To be clear, Don't get me wrong, This is not to say, You could argue/frame this differently but, Some might say... but Problem: AI writing may answer an objection that does not appear in the text. Watch for an unattributed statement about what the writer does not mean, especially when the topic appears nowhere else. A direct claim such as "the API is not thread-safe" is not this pattern. Before:
This isn't mainly about prompt length, and I'm not arguing that documentation doesn't matter. You could categorize the problem another way, but the issue is whether the agent can use the instruction when it acts. After: The issue is whether the agent can use the instruction when it acts.
Remove only the unsupported defense. If it contains a real claim, state that claim directly. Keep an objection when the text names its source or answers it in full.
35. Rejecting fake alternatives
Phrases to watch: A tempting option/approach would be, One might be tempted to, An obvious approach would be, You might think... but, It would be easy to just, Some would suggest Problem: AI writing may introduce an option that no reader would consider, reject it in a clause, and never mention it again. This often leaves an old drafting idea in the final text. Remove the fake option and state the real constraint directly. Before:
Session tokens are rotated every 24 hours. A tempting approach would be to rotate them by restarting the auth service on a cron job, but that would drop every active session. Rotation happens in place, and clients refresh transparently. After: Session tokens are rotated every 24 hours, in place, and clients refresh transparently.
One rejected option may be valid. Several short, unrelated rejections are a stronger sign. Ask what new information each sentence adds. If it only records an earlier edit, rewrite the paragraph around its main point.
36. Over-fragmentation
Signs to watch: Three or more short sentences (under 12 words each) in a row outside of marketing copy. Bullets that have been chopped up into multiple periods. Subordinate clauses converted to standalone sentences. Problem: This pattern is caused by over-applying the other patterns in this skill. When a rewrite strips em dashes, kills -ing phrases, and "varies rhythm" mechanically, the result is staccato - which is its own AI tell. Real human writing breathes; it has subordinate clauses, qualifiers, and connected reasoning across sentence boundaries. Before (over-humanized CV bullet):
Led VPIM campaign. Budget was 12B VND. Team grew to 12 people. Top-3 ranking three years. After: Led the 12B VND VPIM KOL/KOC campaign with a team of 12, contributing to a top-3 social media ranking three years running. Before (over-humanized blog paragraph): AI tools speed up boring code. They are bad at architecture. Tests still matter. You should review every line. After: AI tools speed up the boring parts but stay bad at architecture, which means tests still matter and you still want to review the lines you accept.
Check for false positives
What not to flag
A person may use some of these patterns. Do not treat any item below as proof by itself:
- Perfect grammar and consistent style. Many writers are professionals or have been edited. Polish does not equal AI.
- Mixed casual and formal styles. This can reflect the writer's field, age, or personal habits.
- "Bland" or "robotic" prose. AI prose has specific tells. Generic dryness without those tells is just dry writing.
- Formal or academic words. Pattern 7 lists specific words that AI writing overuses. Do not simplify every formal word.
- Letter-style opening or closing on a comment. Salutations and sign-offs predate ChatGPT by centuries.
- Common transition words in isolation. Additionally, moreover, consequently are AI-coded only when piled up. One however is not a tell.
- Curly quotes alone. macOS, Word, Google Docs, and most CMSes auto-curl by default. Curly quotes only count when stacked with other tells.
- Em dashes alone. Many editors and journalists use them often. Em dashes are evidence only when paired with formulaic sales-y rhythm.
- One short sentence for emphasis. Flag dramatic fragments only when several appear in a row.
- Deliberate repeated openings. Writers may repeat an opening to build rhythm or pressure, as in "She came. She saw. She conquered." Change it only when the repetition adds nothing.
- "Honestly" or "look" mid-sentence. These are ordinary in casual writing. The tell is the standalone theatrical opener, not the word itself.
- Useful limits and disclaimers. Keep scope statements, legal and safety notices, real corrections, named objections, replies, and FAQ answers.
- Real alternatives. Keep options that a reader may consider in a design document, tutorial, or argument. Remove only an unlikely option that the text dismisses and never uses again.
- Unsourced claims. Most of the web is unsourced. Lack of citations doesn't prove anything.
- Correct, complex formatting. Visual editors and templates produce clean output without any AI.
- Secondhand text. Do not rewrite watched phrases inside quotations, titles, proper names, or examples where the phrase is being discussed rather than used.
When unsure, look for several patterns together. One em dash proves nothing. Several stock patterns in the same passage are stronger evidence.
Human details to keep
These details often carry the writer's voice. Keep them unless they hurt the meaning:
- Specific, unusual details. Keep a real address, an odd quote, or a phrase such as "the lawyer who used to work upstairs from my dentist."
- Mixed feelings and unresolved tension. Keep lines such as "I think this is mostly good, but it bothers me, and I can't fully explain why."
- Dated, era-bound references. Slang, memes, or in-jokes that map to a specific year and subculture. Models lag by a year or more.
- Deliberate first-person choices. Keep a cut or word choice when the writer can explain why it belongs.
- Variety in sentence length. Real writing alternates short and long. AI writing tends toward an even, mid-length cadence.
- Genuine asides, parentheticals, or self-corrections. "(I keep wanting to say 'almost' here, but it really was certain.)" Models rarely interrupt themselves like this.
- Edits made before November 30, 2022. ChatGPT's public launch. Anything older than that is, with very rare exceptions, not AI-written.
How to return the result
Pasted text (default).
- Step 0: Identify document type, load the matching profile.
- Read the input text carefully.
- Identify AI patterns relevant to the document type.
- Rewrite each problematic section, preserving format conventions. Apply the em/en dash rule (pattern 14).
- Check for over-fragmentation (pattern 36) - fix any three-short-sentences-in-a-row clusters.
- Ensure the revised text:
- Sounds natural for its format when read aloud
- Varies sentence structure where format allows
- Preserves CV/list/report parallelism where format requires
- Uses specific details over vague claims
- Uses simple constructions (is/are/has) where appropriate
- Present a draft humanized version.
- Self-prompt: "What still reads as AI here?" Answer briefly with remaining tells (if any).
- Self-prompt: "Did the rewrite add or remove any fact, name, number, date, quote, citation, ranking, or other claim?" Treat any unsupported addition or lost claim as an error.
- Present the final version.
Provide:
- Detected document type (one line)
- Draft rewrite
- What still reads as AI here? (brief bullets)
- Final rewrite
- Brief summary of changes made (optional)
File mode. When the user names a file, run the full rewrite process but write only the final text to the file. Change prose only. Keep code blocks, YAML metadata, data, and link targets unchanged. Then give the user a short summary.
Embedded mode. When another task uses this skill for a pull request, commit message, or document, return only the final text.
Vietnamese AI-Tell Patterns
When humanizing Vietnamese content, also check for these Vietnamese-specific AI/machine-translation tells (from content-creator/languages/vi.md): "trong thời đại ngày nay", "trong cuộc sống bận rộn", "đừng ngần ngại", "đừng bỏ lỡ cơ hội này", "hãy cùng khám phá", "đỉnh cao của chất lượng", "giải pháp toàn diện", "đột phá", "tiên phong". For deeper Vietnamese register and platform conventions, load matching files from skills/vietnamese-language/ via its SKILL.md routing table.
Source
This skill is based on Wikipedia: Signs of AI writing, maintained by WikiProject AI Cleanup. Its patterns come from reviews of AI-generated text on Wikipedia.
Wikipedia's main point: "LLMs use statistical algorithms to guess what should come next. The result tends toward the most statistically likely result that applies to the widest variety of cases." The fix is specificity, format-awareness, and connected reasoning - not just chopping sentences shorter.