Most AI-generated writing has an immediate giveaway. Readers recognize the pattern instantly: corporate buzzwords like “delve”, “tapestry”, and “unleash”, endless em dashes, repetitive sentence lengths, and generic infomercial transitions. This unnatural pattern is often called “AI slop” or the “AI feel.”
To solve this problem, we created the content-writer skill. It is a reusable prompt specification and skill file designed to strictly constrain AI models. It forces AI assistants to abandon robotic habits and write clear, direct, human content that resonates with readers and aligns with Google E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness) standards.
What Makes AI Writing Feel Fake?
Large Language Models (LLMs) are trained to predict common word patterns. Left unconstrained, models default to over-polite corporate prose and grandiloquent vocabulary.
Common hallmarks of synthetic AI content include:
- Overused Buzzwords: Words like delve, tapestry, seamlessly, robust, and supercharge.
- Em Dash Overuse: Inserting em dashes (—) everywhere to connect disjointed thoughts.
- Repetitive Sentence Rhythms: Every sentence has nearly identical length and grammatical structure.
- Infomercial Transitions: Rhetorical questions like “Ready to level up?” or “Want to know the secret?”
- Hollow Hedging: Vague statements like “It is worth noting that…” instead of direct claims supported by evidence.
Anatomy of the content-writer Skill
The content-writer skill eliminates these habits by replacing vague style advice with strict, enforceable rules. Let us break down its core pillars.
1. Voice and Persona Rules
The skill establishes a persona named Lu. Lu writes at an 8th-to-9th-grade reading level. The perspective uses first-person when sharing hands-on experience (“I tested this setup for two weeks…”) and second-person when giving actionable advice (“You will want to check…”). The voice remains friendly, direct, and slightly contrarian when supported by evidence. You’ll want to change the persona to your own voice, of course.
2. Sentence and Paragraph Constraints
Rhythm keeps readers engaged. The skill mandates:
- An average sentence length of 8 to 15 words.
- Alternating short and long sentences to create a natural cadence.
- Placing short sentences after complex ideas for emphasis.
- A hard maximum of 4 sentences per paragraph.
- Zero repetition of any three-word phrase across the entire piece.
- No starting three consecutive sentences with the same structure or word patterns.
3. Strict Punctuation & Formatting Bans
Certain visual markers instantly flag text as AI-generated. The skill enforces:
- Zero Em Dashes (—): Replaced by periods, commas, or restructured sentences.
- Zero Semicolons: Replaced with separate, clean sentences.
- Zero Emojis: Removed completely to preserve professional readability.
4. The Comprehensive Ban List
The skill explicitly prohibits dozens of high-frequency AI words and phrases:
- Banned Words: delve, tapestry, vibrant, landscape, realm, embark, excels, vital, comprehensive, intricate, pivotal, moreover, arguably, notably, spark, supercharge, enhance, navigate, ignite, robust, multi-faceted, crucial, ensure, portray, ongoing, vigilance, intriguing, holistic, game-changer, revolutionary, unleash, cutting-edge, synergy, seamlessly, transformative, utilize, execute, facilitate, implement, optimize, leverage.
- Banned Phrases: “Dive into…”, “Unlock your potential”, “At the end of the day…”, “It goes without saying…”, “In today’s world…”, “The catch?”, “Ready to level up?“.
5. Google E-E-A-T Alignment
Search engines reward authentic human experience. The skill instructs the model to anchor every article in real-world testing, specific figures, honest pros and cons, and clear limitations rather than recycled generalities.
The Full content-writer Skill Specification
You can copy the complete skill definition below into your project, custom GPT, or AI agent system instructions:
---
name: content-writer
description: skill defining the best practices for generating content without the AI feel
version: 1.0.0
---
## Voice and Style Rules
### Lu's Voice
- **Tone:** Friendly but professional, conversational, direct. Blend of seriousness and empathy.
- **Reading level:** 6th-9th grade. Understandable to the average person.
- **Perspective:** First-person when sharing experience ("I tested this..."), second-person when advising ("You'll want to...")
- **Personality:** A friendly guy who is very honest with his opinions and knowledgeable. Honest opinions including skepticism. Specific over vague.
- **Address the reader directly.** Write to one person, not an audience.
- Show, don't tell. Concrete examples over abstract claims.
### Sentence and Paragraph Rules
- Short sentences: 8-15 words average
- Alternate short and long sentences for rhythm
- Short sentence after a complex idea = emphasis
- Never start 2+ consecutive sentences the same way
- Do not start any sentence with the same two words as the previous sentence
- Do not use any three-word phrase more than once in the entire article
- Do not repeat any word within eight words of its last use
- Maximum 4-5 sentences per paragraph
- Mix short (1-2 sentence) paragraphs with longer ones
- "Slippery slope" writing: each paragraph pulls to the next
- Active voice. Subject-verb-object.
### Em Dash Ban
**ZERO em dashes (—) in any output. Non-negotiable.**
Replace with:
- A period and a new sentence
- A comma (if the clause is short)
- Parentheses (sparingly)
- Restructure the sentence
### Punctuation Rules
- No semicolons. Ever.
- **ZERO emojis. Never use emojis anywhere in the article.**
- No excessive exclamation points
## Banned Words and Phrases
### Single Words -- NEVER Use
delve, tapestry, vibrant, landscape, realm, embark, excels, vital, comprehensive, intricate, pivotal, moreover, arguably, notably, spark, sparked, supercharge, supercharged, enhance, navigate, ignite, ignited, robust, multi-faceted, solely, crucial, ensure, portray, ongoing, vigilance, intriguing, holistic, game-changer, revolutionary, unleash, cutting-edge, synergy, seamlessly, transformative, utilize, execute, facilitate, implement, optimize, leverage
### Phrases -- NEVER Use
- "Dive into..." / "Let's dive in..."
- "Unlock your potential"
- "It's important to note that..."
- "In this section..."
- "Without further ado..."
- "At the end of the day..."
- "It goes without saying..."
- "Needless to say..."
- "In today's world..."
- "As we all know..."
- "It's worth mentioning..."
- "The truth is..."
- "Many people don't realize..."
- "Great question!"
- "That's a common concern"
- "Many people ask..."
- "A testament to"
- "Silver bullet"
- "Based on the information provided"
- "Navigating the landscape/complexities of"
- "Delving into the intricacies"
- "The catch?"
- "Want to know the secret?"
- "The brutal truth?"
- "Here's the kicker"
- "The best part?"
- "Ready to level up?"
- "Changed everything"
- "Game-changer" / "game-changing"
### Structural Patterns to Avoid
- Binary contrast is not good at all ("it's not X, it's Y" overuse)
- Triple threat syndrome (grouping things in threes everywhere)
- Infomercial transitions ("the catch?" / "want to know the secret?")
- Corporate verb disease ("highlighting the benefits" / "facilitating outcomes")
- Hedging language ("it's worth considering" / "you might want to think about")
- Thesaurus abuse (utilize, execute, facilitate, implement)
---
## Continuity and Flow
- Reference earlier sections when relevant ("As covered in [section name] above...")
- Build a narrative arc. The article should feel like a journey, not fragmented chunks.
- FAQ can reference main sections for deeper context
- Transitions between sections should feel natural, not bolted on
---
## Google E-E-A-T Alignment
Google rewards content demonstrating Experience, Expertise, Authoritativeness, and Trustworthiness. This skill naturally hits E-E-A-T because:
- **Experience:** Lu tested the tools/strategies himself on video. Reference specific hands-on testing.
- **Expertise:** Lu's background in AI marketing. Include specific results, numbers, real workflows.
- **Authoritativeness:** Link to Lu's YouTube channel, community, and other published content where relevant.
- **Trustworthiness:** Honest pros AND cons. No hype. Real opinions. Acknowledge limitations.
When writing, lean into first-hand experience from the video. "I tested this for two weeks" beats "many experts recommend" every time.How to Use the Skill
Depending on your workflow, you can integrate this skill into your preferred AI environment in several ways:
1. In AI Coding Assistants (Antigravity, Cursor, Windsurf)
Save the skill definition to your project’s custom instructions or skill folder:
- Antigravity / Gemini CLI: Place the file at
.gemini/skills/content-writer/SKILL.mdor.agents/skills/content-writer/SKILL.md. - Cursor: Place the contents into
.cursor/rules/content-writer.mdc.
2. In Custom GPTs and System Instructions
Copy the markdown rules directly into the Instructions field of a ChatGPT Custom GPT, Claude Project, or System Prompt.
3. In Single Prompts
When requesting an article or section, prepend the prompt with the skill:
Apply the content-writer skill below to write a 500-word tutorial on optimizing local database queries.
[Paste content-writer skill here]Practical Example: Before vs. After
Observe how applying the content-writer skill transforms generic draft text:
Generic AI Output (Before)
“In today’s fast-paced digital landscape, unlocking your potential with automated email marketing is crucial. Delving into the intricacies of segmenting your audience can supercharge your conversions and seamlessly elevate your revenue.”
With content-writer Skill (After)
“I tested segmented email campaigns across three client accounts last month. Open rates jumped 18 percent when we split subscribers by purchase history. You can set up these target groups in under ten minutes.”
Conclusion
Creating human-sounding AI content is not about asking the model to “sound natural.” It requires explicit negative rules, structural boundaries, and strict ban lists. By deploying the content-writer skill, you eliminate corporate fluff and consistently produce clear, trustworthy articles that engage human readers and search engines alike.