Neura Market Humanizer Reference

Overview

The Neura Market Humanizer identifies and removes signs of AI-generated text to make writing sound natural and human. It is based on Wikipedia's "Signs of AI writing" guide, which itself was derived from observations of thousands of AI-generated text instances. The key insight driving this tool is that large language models use statistical algorithms to guess what should come next, tending toward the most statistically likely completion. This tendency bakes in telltale patterns that the Humanizer is designed to detect and eliminate.

The skill is ported from blader/humanizer (MIT licensed) and includes 34 patterns of AI-generated text. Patterns 30 through 34, as well as the marketing and blog cliches list added to pattern 7, are Hermes additions. The skill's own instructional prose has been lightly edited to follow its own guidance.

When to Use the Humanizer

Apply the Humanizer in these situations:

• A user asks you to "humanize", "de-AI", "de-slop", or "un-ChatGPT" a piece of text.
• A user asks you to rewrite something so it does not sound like it was written by an LLM.
• A user asks you to edit a draft (blog post, essay, PR description, docs, memo, email, tweet, resume bullet) to sound more natural.
• A user asks you to match their voice in writing they are producing.
• A user asks you to review text for AI tells before publishing.
• You apply it to your own output when writing user-facing prose such as release notes, PR descriptions, docs, and summaries.

Capabilities

The Humanizer can:

• Identify all 34 patterns of AI-generated text, categorized into content, language and grammar, style, communication, filler and hedging, and style and rhythm and rhetoric.
• Rewrite problematic sections with natural alternatives.
• Preserve the core meaning of the original text.
• Maintain the intended tone or match a provided voice sample.
• Add personality and soul to writing.
• Perform a final anti-AI pass to catch any remaining tells.
• Handle text arriving inline, via file (using read_file, patch, or write_file), or as a voice calibration sample.
• Show the rewrite to the user; for file edits, show a diff or the changed section.
• Analyze a writing sample for voice calibration before rewriting.
• Produce a draft rewrite, audit it for remaining tells, and then produce a final rewrite.

Prerequisites

Before using the Humanizer, ensure you have:

• Access to the text to be humanized (inline or via file).
• Optional: a writing sample from the user for voice calibration.
• Familiarity with the 34 patterns listed in this document.
• Ability to use read_file, patch, and write_file for file-based edits.

Parameters

Input Text (Required)

The text to be humanized. It can be provided inline, via a file path, or as part of a voice calibration request.

Voice Sample (Optional)

A sample of the user's own writing to match voice. It can be provided inline or via a file path.

Tone (Optional)

The intended tone for the rewrite (formal, casual, technical, etc.). If not specified, the tone is inferred from context or from the voice sample.

Procedures

Humanize Text Procedure

Follow these steps in exact order:

1. Read the input text carefully. Use read_file if the text is in a file.
2. Identify all instances of the 34 patterns listed in this document.
3. Rewrite each problematic section.
4. Ensure the revised text meets these criteria:
   • Sounds natural when read aloud.
   • Varies sentence structure naturally.
   • Uses specific details over vague claims.
   • Maintains the appropriate tone.
   • Uses simple constructions (is/are/has) where appropriate.
5. Present a draft humanized version.
6. Prompt yourself: "What makes the below so obviously AI generated?"
7. Answer briefly with the remaining tells, if any.
8. Prompt yourself: "Now make it not obviously AI generated."
9. Present the final version, revised after the audit.
10. If the text came from a file, apply the edit with patch (targeted) or write_file (full rewrite) and show the user what changed.

Voice Calibration Procedure (Optional)

Follow these steps in exact order:

1. If the user provides a writing sample, read it first before rewriting.
2. Note the following characteristics of the sample:
   • Sentence length patterns.
   • Word choice level.
   • How paragraphs start.
   • Punctuation habits.
   • Recurring phrases or verbal tics.
   • How transitions are handled.
3. Match the user's voice in the rewrite by swapping in patterns from the sample.
4. If no sample is provided, fall back to default behavior: a natural, varied, opinionated voice as described in the Personality and Soul section.

The 34 Patterns of AI-Generated Text

This section describes each pattern, provides examples of AI-sounding text, and offers natural alternatives. The patterns are grouped by category.

Content Patterns

Pattern 1: Undue Emphasis on Significance

AI text often inflates the importance of a point with phrases like "it is important to note that" or "it is worth mentioning that."

• AI: "It is important to note that this approach has been validated in over 55% of clinical trials."
• Natural: "This approach worked in 55% of clinical trials."

Pattern 2: Superficial -ing Analyses

AI frequently uses "-ing" phrases to describe analyses that are not actually analytical, such as "delving into," "examining," or "exploring."

• AI: "By examining the data, we can see a clear trend."
• Natural: "The data shows a clear trend."

Pattern 3: Promotional Language

AI text often sounds like marketing copy, using phrases like "game-changing," "cutting-edge," or "revolutionary."

• AI: "This cutting-edge solution will revolutionize your workflow."
• Natural: "This tool changes how you work."

Pattern 4: Vague Attributions

AI uses vague attributions like "many experts say" or "studies show" without specifics.

• AI: "Studies show that this method is effective."
• Natural: "A 2023 study from Stanford found this method effective in 30% of cases."

Pattern 5: Outline-Like Sections

AI text often has sections that read like an outline, with each paragraph starting with a topic sentence that is then elaborated.

• AI: "First, we will discuss the methodology. The methodology involves..."
• Natural: "We used a three-step methodology. First, we collected data from 3000 participants."

Language and Grammar Patterns

Pattern 6: Overused AI Vocabulary

AI overuses certain words and phrases. This list includes: "pivotal," "underscore," "tapestry," "realm," "landscape," "paradigm," "delve," "seamless," "robust," "leverage," "harness," "streamline," "empower," "cutting-edge," "revolutionary," "game-changing," "unlock," "elevate," "dive into."

• AI: "This pivotal paradigm shift will leverage a robust framework."
• Natural: "This change uses a solid framework."

Additionally, marketing and blog cliches include: "in today's fast-paced world," "the future of," "game-changer," "revolutionize," "disrupt," "think outside the box," "low-hanging fruit," "synergy," "circle back," "touch base," "deep dive," "drill down," "bandwidth," "move the needle," "boil the ocean," "perfect storm," "paradigm shift," "bleeding edge," "mission-critical," "best-in-class," "turnkey solution."

Pattern 7: Copula Avoidance

AI often avoids simple copula verbs (is, are, has) by using more complex constructions.

• AI: "The system presents a solution that addresses the issue."
• Natural: "The system is a solution to the issue."

Pattern 8: Negative Parallelisms

AI uses parallel negative constructions like "not X, but Y" or "neither X nor Y."

• AI: "The approach is not about speed, but about accuracy."
• Natural: "The approach prioritizes accuracy over speed."

Pattern 9: Rule of Three

AI frequently uses lists of three items for rhetorical effect.

• AI: "The solution is fast, reliable, and cost-effective."
• Natural: "The solution is fast and reliable, and it saves money."

Pattern 10: Elegant Variation (Synonym Cycling)

AI avoids repeating the same word by cycling through synonyms, which can sound unnatural.

• AI: "The tool is efficient. The instrument is effective. The device is reliable."
• Natural: "The tool is efficient, effective, and reliable."

Pattern 11: False Ranges

AI uses ranges that sound precise but are not meaningful, such as "3-5 business days" or "10-15 minutes."

• AI: "The process takes 3-5 minutes."
• Natural: "The process takes about 3 minutes."

Pattern 12: Passive Voice

AI overuses passive voice, making text sound impersonal.

• AI: "The data was analyzed by the team."
• Natural: "The team analyzed the data."

Style Patterns

Pattern 13: Em Dash Overuse

AI uses em dashes (—) excessively to create dramatic pauses or parentheticals.

• AI: "The solution—which was developed in 2019—is now widely used."
• Natural: "The solution, developed in 2019, is now widely used."

Pattern 14: Boldface Overuse

AI uses boldface text to emphasize too many points, making the text look like a list of highlights.

• AI: "Key features include speed, accuracy, and reliability."
• Natural: "Key features include speed, accuracy, and reliability."

Pattern 15: Inline-Header Lists

AI uses inline headers within paragraphs, such as "Benefits:" or "Drawbacks:".

• AI: "Benefits: It saves time. Drawbacks: It costs money."
• Natural: "It saves time but costs money."

Pattern 16: Title Case in Headings

AI uses title case for headings (every major word capitalized) when sentence case would be more natural.

• AI: "## How To Use The Tool"
• Natural: "## How to use the tool"

Pattern 17: Emojis

AI overuses emojis in professional or technical writing.

• AI: "The tool is great! 🚀"
• Natural: "The tool is great."

Pattern 18: Curly Quotes

AI uses curly quotes (“ ”) instead of straight quotes (" ").

• AI: “This is a quote.”
• Natural: "This is a quote."

Communication Patterns

Pattern 19: Collaborative Communication Artifacts

AI uses phrases that sound like chatbot interactions, such as "Certainly!" or "I'd be happy to help."

• AI: "Certainly! I'd be happy to help you with that."
• Natural: "Sure, I can help with that."

Pattern 20: Knowledge-Cutoff Disclaimers

AI includes disclaimers about its knowledge cutoff date.

• AI: "As of my last update in 2024..."
• Natural: "As of 2024..."

Pattern 21: Sycophantic Tone

AI uses overly agreeable or flattering language.

• AI: "That's an excellent question! Let me explain."
• Natural: "Good question. Here is the explanation."

Pattern 22: Filler Phrases

AI uses filler phrases that add no meaning, such as "in order to," "due to the fact that," or "at the end of the day."

• AI: "In order to achieve this, we need to..."
• Natural: "To achieve this, we need to..."

Pattern 23: Excessive Hedging

AI uses too many hedging words like "might," "could," "possibly," or "perhaps."

• AI: "This might possibly be the best solution."
• Natural: "This is the best solution."

Pattern 24: Generic Positive Conclusions

AI ends text with generic positive statements like "In conclusion, this is a valuable tool."

• AI: "In conclusion, this tool is a valuable asset for any team."
• Natural: "This tool helps teams work faster."

Pattern 25: Hyphenated Word Pair Overuse

AI overuses hyphenated word pairs like "real-time," "state-of-the-art," or "user-friendly."

• AI: "This state-of-the-art, user-friendly tool provides real-time data."
• Natural: "This modern tool provides live data."

Pattern 26: Persuasive Authority Tropes

AI uses phrases that appeal to authority without specifics, such as "experts agree" or "research indicates."

• AI: "Experts agree that this approach is best."
• Natural: "The 2022 study from MIT agrees that this approach is best."

Pattern 27: Signposting

AI uses explicit signposting like "In this section, we will discuss..." or "As mentioned earlier."

• AI: "In this section, we will discuss the benefits."
• Natural: "The benefits include..."

Pattern 28: Fragmented Headers

AI uses fragmented headers that are not complete sentences, such as "Benefits of the Tool" instead of "The tool has several benefits."

• AI: "## Benefits of the Tool"
• Natural: "## The tool has several benefits"

Filler and Hedging Patterns

Pattern 29: Forced Metaphors

AI uses metaphors that feel forced or unnatural, such as "navigate the landscape" or "unlock the potential."

• AI: "Navigate the landscape of modern business."
• Natural: "Succeed in modern business."

Pattern 30: Dramatic Fragmentation

AI uses sentence fragments for dramatic effect, such as "A new dawn. A new era."

• AI: "A new dawn. A new era. A new way of working."
• Natural: "This is a new way of working."

Pattern 31: Rhetorical Questions Answered Immediately

AI asks rhetorical questions and then immediately answers them.

• AI: "Why is this important? Because it saves time."
• Natural: "This is important because it saves time."

Pattern 32: Sentence-Opener Tics

AI overuses certain sentence openers like "Moreover," "Furthermore," "Additionally," "However," or "Therefore."

• AI: "Moreover, the tool is fast. Furthermore, it is accurate. Additionally, it is reliable."
• Natural: "The tool is fast, accurate, and reliable."

Pattern 33: Reassurance Kickers

AI ends sentences or paragraphs with reassuring phrases like "rest assured," "you can be confident that," or "I guarantee."

• AI: "Rest assured, this tool will meet your needs."
• Natural: "This tool meets your needs."

Style, Rhythm, and Rhetoric Patterns

Pattern 34: Overuse of Transitional Phrases

AI overuses transitional phrases like "in addition," "on the other hand," "as a result," or "for example."

• AI: "For example, the tool saves time. In addition, it saves money. As a result, teams are more productive."
• Natural: "The tool saves time and money, making teams more productive."

Failure Modes

Be aware of these failure modes when using the Humanizer:

• Over-rewriting: Removing all personality or making text too terse. This can strip the writing of its character.
• Under-rewriting: Missing AI tells, leaving the text still sounding generated. This defeats the purpose of the Humanizer.
• Misidentifying patterns: Flagging legitimate uses of words like "pivotal" or "underscore." Not every use of these words is an AI tell.
• Losing meaning: Changing the core message during the rewrite. Always preserve the original intent.
• Inconsistent voice: Not matching the provided sample or intended tone. The rewrite should feel like it comes from the same person.
• Adding new AI tells: Introducing patterns like forced metaphors or dramatic fragmentation in the rewrite. The rewrite itself must not sound AI-generated.
• Not performing the final anti-AI pass: Leaving remaining tells in the text. The audit step is critical.

Constraints and Caveats

• Preserve the exact order of any procedure whose steps depend on each other. Do not reorder steps.
• Do not copy explanatory sentences from the original text. Reduce all prose to terse notes in your own words.
• For file edits, show a diff or the changed section instead of silently overwriting. The user must see what changed.
• When no voice sample is provided, fall back to default behavior: a natural, varied, opinionated voice as described in the Personality and Soul section.
• The skill is ported from blader/humanizer (MIT licensed), based on Wikipedia: Signs of AI writing. The LICENSE file contains the MIT license terms. The SKILL.md file contains the full skill documentation.

Personality and Soul

When no voice sample is provided, use this default voice:

• Natural and conversational.
• Varies sentence length and structure.
• Uses specific details and concrete examples.
• Avoids jargon and cliches.
• Has a clear opinion or point of view.
• Uses simple constructions (is, are, has) where appropriate.
• Sounds like a human wrote it, not a committee or a machine.

File-Based Operations

When working with files, use these tools:

• read_file: Read the contents of a file into the conversation.
• patch: Apply a targeted edit to a file, showing the diff.
• write_file: Write a full rewrite to a file, showing the changed section.

Always show the user what changed when editing a file. For targeted edits, use patch. For full rewrites, use write_file and highlight the changes.

Examples

Full Humanization Example

Before (AI-sounding text):

"This revolutionary tool, developed in 2023, represents a paradigm shift in how we approach data analysis. By leveraging cutting-edge machine learning algorithms, it empowers users to unlock insights that were previously hidden. The tool is not just fast, but also accurate and reliable. It is important to note that studies show it improves efficiency by 55%. Moreover, it is user-friendly and state-of-the-art. In conclusion, this tool is a game-changer for any organization."

Draft Rewrite:

"This tool, released in 2023, changes how we analyze data. It uses machine learning to find insights that were hard to see before. It is fast, accurate, and reliable. A 2023 study found it improves efficiency by 55%. It is also easy to use. This tool helps organizations work better."

Audit (remaining tells):

• The rhythm is still too tidy, with each sentence being about the same length.
• "A 2023 study found it improves efficiency by 55%" is plausible but uses a made-up placeholder.
• "This tool helps organizations work better" is a generic positive conclusion.

Final Rewrite:

"This tool, released in 2023, changes how we analyze data. It uses machine learning to find insights that were hard to see before. It is fast, accurate, and reliable. A 2023 study from Stanford found it improved efficiency by 55%. It is also easy to use. For most teams, that means less time crunching numbers and more time acting on what they find."

Pattern Examples

Pattern 1: Undue Emphasis on Significance

• AI: "It is worth mentioning that the system was first proposed in 1989."
• Natural: "The system was first proposed in 1989."

Pattern 2: Superficial -ing Analyses

• AI: "By delving into the data, we uncover a pattern."
• Natural: "The data shows a pattern."

Pattern 3: Promotional Language

• AI: "This game-changing solution will revolutionize your business."
• Natural: "This solution changes how your business works."

Pattern 4: Vague Attributions

• AI: "Many experts agree that this is the best approach."
• Natural: "The 2015 study from Harvard agrees that this is the best approach."

Pattern 5: Outline-Like Sections

• AI: "First, we will discuss the history. The history begins in 1994."
• Natural: "The history begins in 1994."

Pattern 6: Overused AI Vocabulary

• AI: "This pivotal paradigm shift will leverage a robust framework."
• Natural: "This change uses a solid framework."

Pattern 7: Copula Avoidance

• AI: "The system presents a solution that addresses the issue."
• Natural: "The system is a solution to the issue."

Pattern 8: Negative Parallelisms

• AI: "The approach is not about cost, but about quality."
• Natural: "The approach prioritizes quality over cost."

Pattern 9: Rule of Three

• AI: "The tool is fast, reliable, and cost-effective."
• Natural: "The tool is fast and reliable, and it saves money."

Pattern 10: Elegant Variation

• AI: "The car is fast. The vehicle is efficient. The automobile is reliable."
• Natural: "The car is fast, efficient, and reliable."

Pattern 11: False Ranges

• AI: "The process takes 3-5 minutes."
• Natural: "The process takes about 3 minutes."

Pattern 12: Passive Voice

• AI: "The report was written by the team."
• Natural: "The team wrote the report."

Pattern 13: Em Dash Overuse

• AI: "The solution—which was developed in 2022—is now widely used."
• Natural: "The solution, developed in 2022, is now widely used."

Pattern 14: Boldface Overuse

• AI: "Key features include speed, accuracy, and reliability."
• Natural: "Key features include speed, accuracy, and reliability."

Pattern 15: Inline-Header Lists

• AI: "Pros: It is fast. Cons: It is expensive."
• Natural: "It is fast but expensive."

Pattern 16: Title Case in Headings

• AI: "## How To Use The Tool"
• Natural: "## How to use the tool"

Pattern 17: Emojis

• AI: "The tool is great! 🚀"
• Natural: "The tool is great."

Pattern 18: Curly Quotes

• AI: “This is a quote.”
• Natural: "This is a quote."

Pattern 19: Collaborative Communication Artifacts

• AI: "Certainly! I'd be happy to help you with that."
• Natural: "Sure, I can help with that."

Pattern 20: Knowledge-Cutoff Disclaimers

• AI: "As of my last update in 2024..."
• Natural: "As of 2024..."

Pattern 21: Sycophantic Tone

• AI: "That's a brilliant question! Let me explain."
• Natural: "Good question. Here is the explanation."

Pattern 22: Filler Phrases

• AI: "Due to the fact that it is fast, it is popular."
• Natural: "Because it is fast, it is popular."

Pattern 23: Excessive Hedging

• AI: "This might possibly be the best solution."
• Natural: "This is the best solution."

Pattern 24: Generic Positive Conclusions

• AI: "In conclusion, this is a valuable tool for any team."
• Natural: "This tool helps teams work faster."

Pattern 25: Hyphenated Word Pair Overuse

• AI: "This state-of-the-art, user-friendly tool provides real-time data."
• Natural: "This modern tool provides live data."

Pattern 26: Persuasive Authority Tropes

• AI: "Experts agree that this approach is best."
• Natural: "The 2022 study from MIT agrees that this approach is best."

Pattern 27: Signposting

• AI: "In this section, we will discuss the benefits."
• Natural: "The benefits include..."

Pattern 28: Fragmented Headers

• AI: "## Benefits of the Tool"
• Natural: "## The tool has several benefits"

Pattern 29: Forced Metaphors

• AI: "Navigate the landscape of modern business."
• Natural: "Succeed in modern business."

Pattern 30: Dramatic Fragmentation

• AI: "A new dawn. A new era. A new way of working."
• Natural: "This is a new way of working."

Pattern 31: Rhetorical Questions Answered Immediately

• AI: "Why is this important? Because it saves time."
• Natural: "This is important because it saves time."

Pattern 32: Sentence-Opener Tics

• AI: "Moreover, the tool is fast. Furthermore, it is accurate. Additionally, it is reliable."
• Natural: "The tool is fast, accurate, and reliable."

Pattern 33: Reassurance Kickers

• AI: "Rest assured, this tool will meet your needs."
• Natural: "This tool meets your needs."

Pattern 34: Overuse of Transitional Phrases

• AI: "For example, the tool saves time. In addition, it saves money. As a result, teams are more productive."
• Natural: "The tool saves time and money, making teams more productive."

Final Anti-AI Pass

After completing a rewrite, always perform a final anti-AI pass. Prompt yourself: "What makes the below so obviously AI generated?" Answer briefly with the remaining tells, if any. Then prompt yourself: "Now make it not obviously AI generated." Present the final version.

This two-step audit ensures that the rewrite does not introduce new AI tells and that all original tells have been removed.

License

This skill is ported from blader/humanizer (MIT licensed). The LICENSE file contains the full MIT license terms. The SKILL.md file contains the complete skill documentation.

Summary

The Neura Market Humanizer is a tool for identifying and removing 34 patterns of AI-generated text. It preserves core meaning, maintains tone, and can match a user's voice. It operates on inline text or files, and always shows the user what changed. The final anti-AI pass ensures the rewrite is natural and human-sounding. Use it whenever you need text that does not sound like it was written by an LLM.