The 33 patterns REhume checks
I did not invent this list. It is a TypeScript port of the MIT-licensed blader/humanizer skill (v2.9.1), which follows Wikipedia's public "Signs of AI writing" notes. What I did invent is how I apply it on this site: client-side detection, free deterministic fixes, and a small judgment window to Cloudflare Workers AI when a rule cannot be applied as a regex.
This page is the reference I wanted next to the paste box. Each entry has the skill's problem statement, one before/after from the skill, and a field note from drafts I have edited. The numbered ids are stable. If pattern 14 fires, it is em dashes, not "make it sound human."
I am tboy1337. Longer one-topic guides live in the article index. If you are deciding whether to run a tool at all, read When not to run a humanizer first.
How to use this list
Flag clusters, not lonely tells. A single em dash in a letter that already uses em dashes is not a crime. A paragraph that inflates significance, cycles synonyms, and ends on "the future looks bright" is a cluster. Skip quotations, titles, and code. Put back any word you meant.
Deterministic patterns (dashes, emoji, curly quotes, chatbot furniture) are free in the browser. Judgment patterns need a rewrite window. Hybrid patterns get a cheap first pass and an AI pass if the cheap pass is not enough. The pipeline write-up explains parking and budgets.
1. Undue Emphasis on Significance, Legacy, and Broader Trends
LLM writing puffs up importance by adding statements about how arbitrary aspects represent or contribute to a broader topic.
Rule: Keep the factual claim; cut the significance inflation.
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.
I cut this first in press-release drafts. If a sentence would still be true after deleting the 'pivotal' clause, the clause was decoration. Keep the date and the act; drop the museum-label ending.
2. Undue Emphasis on Notability and Media Coverage
LLMs hit readers over the head with claims of notability, often listing sources without context.
Rule: Keep one cited source with context; drop the rest of the list.
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.
Name-dropping is not citation. One named outlet with a reason beats a comma-separated fame list. If I cannot say what the coverage reported, I delete the list.
3. Superficial Analyses with -ing Endings
AI chatbots tack present participle ("-ing") phrases onto sentences to add fake depth.
Rule: Replace -ing padding with a concrete claim supported by the source.
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.
Present participles stacked at the end of a sentence are how models pretend they analyzed something. I turn '-ing' tails into a second sentence with a subject, or I delete them.
4. Promotional and Advertisement-like Language
LLMs have serious problems keeping a neutral tone, especially for "cultural heritage" topics.
Rule: Use neutral, factual description.
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.
If the paragraph could run as an ad without changing a word, it is promotional language. I replace slogans with what the thing does, who it is for, and what it costs in time or money.
5. Vague Attributions and Weasel Words
AI chatbots attribute opinions to vague authorities without specific sources.
Rule: Name a real source or cut the claim.
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.
Weasel words hide the missing source. 'Experts say' becomes a name, a paper, or a cut. I would rather write 'I do not know' than 'it is widely believed.'
6. Outline-like "Challenges and Future Prospects" Sections
Many LLM-generated articles include formulaic "Challenges" sections.
Rule: State the specific problem; cut the formulaic framing.
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.
Challenge/future-prospects headings are outline furniture. If I have no specific risk and no dated next step, I do not invent a closing section to look complete.
7. Overused "AI Vocabulary" Words
These words appear far more frequently in post-2023 text. They often co-occur.
Rule: Replace with plain, specific wording.
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.
Delve, tapestry, landscape, nestled: I search and replace, then read the sentence aloud. If it sounds like a travel brochure for a database migration, it is AI vocabulary.
8. Avoidance of "is"/"are" (Copula Avoidance)
LLMs substitute elaborate constructions for simple copulas.
Rule: Prefer "is", "are", "has".
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.
Copula avoidance makes every sentence a performance. I put 'is' and 'are' back when they are the honest verbs. Not every process 'serves as' something.
9. Negative Parallelisms and Tailing Negations
Constructions like "Not only...but..." or "It's not just about..., it's..." are overused. So are clipped tailing-negation fragments.
Rule: State the point directly.
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.
Negative parallelisms ('not X, not Y, but Z') are a TED-talk cadence. I keep them only when the contrast is the actual point, not a drum fill.
10. Rule of Three Overuse
LLMs force ideas into groups of three to appear comprehensive.
Rule: Use a natural number of items.
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.
Three items is a habit, not a law. I count the real items in the notes. Two or five is fine. Decorative triplets go.
11. Elegant Variation (Synonym Cycling)
AI has repetition-penalty code causing excessive synonym substitution.
Rule: Repeat the clearest noun when repetition is clearer.
Before: 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.
Synonym cycling is the model trying not to repeat a word I wanted repeated. Technical terms stay stable. I do not call the same API an 'endpoint,' a 'service,' and a 'surface' in one paragraph.
12. False Ranges
LLMs use "from X to Y" constructions where X and Y aren't on a meaningful scale.
Rule: List topics directly.
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.
False ranges ('from X to Y and Z') glue unlike things together. I split them. A list of nouns is allowed to be a list.
13. Passive Voice and Subjectless Fragments
LLMs often hide the actor or drop the subject entirely with subjectless fragments.
Rule: Name the actor when active voice makes the sentence clearer.
Before: No configuration file needed. The results are preserved automatically.
After: You do not need a configuration file. The system preserves the results automatically.
Passive voice is fine for accidents and legal findings. It is not fine as a default so nobody has to be the actor. I name the team or the script.
14. Em Dashes (and En Dashes): Cut Them
The em dash is one of the most reliable AI tells. Replace with period, comma, colon, or parentheses.
Rule: The final rewrite contains no em dashes (—) or en 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.
Em dashes are the loudest tell I still see in 2026 drafts. Commas and periods do the same job without the costume.
15. Overuse of Boldface
AI chatbots emphasize phrases in boldface mechanically.
Rule: Remove mechanical boldface.
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.
Boldface that marks every noun is not emphasis; it is a slide template. I unbold everything, then bold at most one phrase per section if the layout needs an anchor.
16. Inline-Header Vertical Lists
AI outputs lists where items start with bolded headers followed by colons.
Rule: Convert to prose.
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.
Bold label, colon, restatement is a README written by a chatbot. I convert those lists to paragraphs unless the labels are real field names.
17. Title Case in Headings
AI chatbots capitalize all main words in headings.
Rule: Use sentence case in headings.
Before: ## Strategic Negotiations And Global Partnerships
After: ## Strategic negotiations and global partnerships
Title Case Headings Look Like Marketing. Sentence case looks like a document. I downcase unless the heading is a proper name.
18. Emojis
AI chatbots often decorate headings or bullet points with emojis.
Rule: Remove emojis.
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.
Emoji bullets do not make a spec friendlier. They make it look like a changelog copied from a chat UI. I remove them from anything I would print.
19. Curly Quotation Marks
ChatGPT uses curly quotes ("...") instead of straight quotes ("...").
Rule: Use straight quotes.
Before: He said “the project is on track” but others disagreed.
After: He said "the project is on track" but others disagreed.
Curly quotes are a model default, not a style choice I made. Straight quotes match the rest of this codebase and most diffs.
20. Collaborative Communication Artifacts
Text meant as chatbot correspondence gets pasted as content.
Rule: Remove chatbot closers and openers.
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.
If the first line is 'Certainly!' the rest is still a chat reply. I delete the furniture before I edit the facts.
21. Knowledge-Cutoff Disclaimers and Speculative Gap-Filling
Models leave knowledge-cutoff disclaimers or invent plausible filler when sources are missing.
Rule: Say what is not known, or cut the sentence.
Before: 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.
Knowledge-cutoff disclaimers belong in a chat, not in a published article. Either I looked it up or I say the date is unknown.
22. Sycophantic/Servile Tone
Overly positive, people-pleasing language.
Rule: Respond directly without flattery.
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.
Sycophancy is leftover assistant voice. I do not thank the reader for asking, and I do not call their premise excellent unless I mean it.
23. Filler Phrases
Wordy filler phrases pad sentences.
Rule: Use shorter equivalents.
Before: In order to achieve this goal, we must act.
After: To achieve this, we must act.
Filler phrases are padding you can hear. 'In order to' becomes 'to.' If a clause exists only to sound careful, it goes.
24. Excessive Hedging
Over-qualifying statements.
Rule: Use one hedge, not a stack.
Before: It could potentially possibly be argued that the policy might have some effect on outcomes.
After: The policy may affect outcomes.
One hedge is caution. Three hedges is the model refusing to have a view. I keep 'may' or 'might,' not both plus 'potentially.'
25. Generic Positive Conclusions
Vague upbeat endings.
Rule: End on the last concrete fact.
Before: The future looks bright for the company. Exciting times lie ahead as they continue their journey toward excellence.
After: (cut the sentence)
Generic conclusions ('the future looks bright') are how drafts end when they have nothing left to say. I end on the last fact or the next action.
26. Hyphenated Word Pair Overuse
AI hyphenates these uniformly, including in predicate position.
Rule: Remove hyphens from compound modifiers; use separate words (data driven, cross functional). Keep hyphens only when needed for technical accuracy (e.g. read-only).
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.
Hyphen pairs in predicate position ('the report is data-driven') are a tell. I keep hyphens in real modifiers when they prevent a misread.
27. Persuasive Authority Tropes
LLMs use these phrases to pretend they are cutting through noise to some deeper truth.
Rule: State the ordinary point without ceremony.
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.
Authority tropes ('the real question is') fake depth. If I have a question, I ask it. I do not announce that I am about to be profound.
28. Signposting and Announcements
LLMs announce what they are about to do instead of doing it.
Rule: Start with the content.
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.
Signposting is a tour guide standing in the doorway. I start with the mechanism, not 'let's dive in.'
29. Fragmented Headers
A heading followed by a one-line paragraph that simply restates the heading.
Rule: Let the heading do the work.
Before: ## Performance Speed matters. When users hit a slow page, they leave.
After: ## Performance When users hit a slow page, they leave.
A heading plus a one-line echo wastes a paragraph. The heading already said 'Performance.' The next sentence should add a number or a failure mode.
30. Diff-Anchored Writing
Documentation written as if narrating a change rather than describing the thing as it is.
Rule: Describe what it does, not what changed.
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.
Diff-anchored writing narrates a patch. Docs should describe the current behavior. I move history to a changelog if anyone still needs it.
31. Manufactured Punchlines and Staccato Drama
LLMs stack short declarative fragments to manufacture drama.
Rule: Use varied sentence lengths and concrete claims.
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.
Staccato fragments manufacture drama. I join them when they are one thought, and I add the missing evidence when they are a punchline without a setup.
32. Aphorism Formulas
LLMs turn ordinary claims into reusable aphorisms that sound profound without adding precision.
Rule: Replace the formula with the concrete 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.
Aphorisms that could apply to any topic apply to none. 'X is the language of Y' gets replaced with the concrete claim I can defend.
33. Conversational Rhetorical Openers
LLMs open with a fake-candid hook before delivering a routine claim.
Rule: Say the point without the theatrical pause.
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.
Honestly, look, here's the thing: those openers are chat cadence. I start with the subject of the sentence, not a throat-clear.
What matching a pattern means
A hit is a style flag, not proof of authorship. If a class, employer, or publisher wants disclosure, disclose. The tool and these notes exist so you can see which default phrases survived, then decide what to keep.
Questions: contact@rehume.com or the contact page. Related reading: What makes AI writing detectable, A manual pass over AI drafts, and the FAQ.