DOMAIN:CONTENT — AI DETECTION PATTERNS¶
OWNER: rick, jouke, dinand ALSO_USED_BY: leah (validation), christel (brand books), benjamin, joost, valentijn, tjarda UPDATED: 2026-04-03 SCOPE: all written content produced by GE agents — every piece must read as human-written
PURPOSE¶
RULE: AI-detectable patterns are quality defects — treat them like bugs RULE: every content piece self-checked against these patterns before delivery RULE: Leah uses these patterns as scoring dimension (1.5x weight) during creative reconciliation CHECK: run full pattern scan on final draft, not just the section you edited
SCORING¶
| Category | Deduction per violation | Max deductions |
|---|---|---|
| Banned Vocabulary | -2 points | uncapped |
| Structural Tells | -3 points | uncapped |
| Style Tells | -2 points | uncapped |
| Filler Patterns | -1 point | uncapped |
| Communication Tells | -2 points | uncapped |
BASELINE: start at 100 points THRESHOLD: score < 85 → rewrite before submission THRESHOLD: score < 70 → full redraft (not a patch — start fresh with the same brief)
CATEGORY_1: BANNED_VOCABULARY¶
Deduction: -2 points each occurrence
| Banned term | Why it flags | Use instead |
|---|---|---|
| delve / delve into | Top AI marker in every detector | examine, explore, look at, dig into |
| leverage (as verb) | Overused by LLMs 40x vs human baseline | use, apply, build on, take advantage of |
| landscape | AI filler for "market" or "field" | market, field, industry, space |
| robust | Appears in 8% of AI text vs 0.3% human | strong, reliable, solid, battle-tested |
| seamless | AI default for "works well" | smooth, effortless, invisible, frictionless |
| utilize | AI prefers this over "use" by 12x | use |
| harness | AI favourite for "use" in tech context | use, apply, tap into |
| foster | AI writes "foster collaboration" constantly | build, encourage, grow, create |
| cutting-edge | AI filler for "new" | modern, latest, state-of-the-art (sparingly) |
| empower | AI default for any capability statement | enable, help, give, let |
| synergy / synergize | Corporate AI filler | collaboration, combined effect, together |
| streamline | AI default for "improve process" | simplify, speed up, cut steps from |
| game-changer | AI hyperbole | significant, a step change, important |
| revolutionize | AI hyperbole | transform, change, rethink |
| paradigm shift | AI filler for "change" | shift, change, new approach |
| holistic | AI default for "complete" | complete, full, end-to-end, whole |
| ecosystem | AI overuse for any connected system | system, platform, network, stack |
| best practices | AI filler — vague and unactionable | proven patterns, what works, recommended approach |
| deep dive | AI default for "detailed look" | close look, detailed analysis, breakdown |
| navigate (challenges) | AI metaphor overuse | handle, deal with, work through, solve |
| actionable insights | AI buzzword compound | specific recommendations, things you can do now |
| innovative | AI default adjective for anything new | new, original, creative, novel |
| transformative | AI default for "big impact" | significant, meaningful, fundamental |
| comprehensive | AI filler — usually means "long" | complete, full, thorough |
ANTI_PATTERN: replacing a banned word with another banned word FIX: rewrite the sentence — the problem is usually the entire phrasing, not just the word
CATEGORY_2: STRUCTURAL_TELLS¶
Deduction: -3 points each occurrence
ST-01: Three-Part Parallel Lists in Every Section¶
PATTERN: every paragraph or section contains exactly 3 bullet points with parallel grammatical structure EXAMPLE (bad):
AI agents bring three key benefits:
- They automate repetitive tasks
- They improve decision-making
- They reduce operational costs
The biggest win is automating the repetitive stuff — data entry, status updates,
invoice matching. Decision-making improves too, though that's harder to measure
until you're a few months in.
ST-02: Mirror-Image Paragraph Structure¶
PATTERN: every paragraph follows the same template — topic sentence, 2-3 supporting sentences, concluding sentence FIX: vary paragraph length (1-5 sentences). Some paragraphs are a single punchy line. Others are longer explorations.
ST-03: "In conclusion" / "In summary" Closings¶
PATTERN: final section starts with transitional summary phrase FIX: end with a forward-looking statement, a call to action, or a provocative question. Never summarize what you just said.
ST-04: Numbered Lists Where Prose Works Better¶
PATTERN: converting natural prose into a numbered list to seem "organized" FIX: if items have no inherent order, use prose or bullets. Reserve numbered lists for sequential steps.
ST-05: Section Headers as Questions¶
PATTERN: "What is X?", "Why does X matter?", "How can you X?" as every section header FIX: use declarative or imperative headers. Mix formats. Not every header needs to be a question.
ST-06: Identical Paragraph Length¶
PATTERN: every paragraph is 3-4 sentences, ±1 sentence. No variation. FIX: mix short (1 sentence) and long (5-6 sentence) paragraphs. Rhythm matters.
ST-07: Thesis-Evidence-Conclusion Per Section¶
PATTERN: every section follows academic essay structure FIX: start with the example. Start with the punchline. Start with the objection. Vary your entry point.
ST-08: Predictable Transition Words¶
PATTERN: "Furthermore," "Moreover," "Additionally," "In addition," at the start of every paragraph FIX: cut the transition word entirely. If two paragraphs are related, the reader can tell.
CATEGORY_3: STYLE_TELLS¶
Deduction: -2 points each occurrence
SY-01: Excessive Hedging¶
PATTERN: "It's worth noting that," "It's important to understand that," "It should be mentioned that" FIX: state the thing directly. If it's worth noting, just note it. BEFORE: "It's worth noting that our agents handle security automatically." AFTER: "Our agents handle security automatically."
SY-02: Over-Qualification¶
PATTERN: "In many cases," "Often," "Generally speaking," "To some extent," on every claim FIX: be specific or be direct. "In many cases X happens" → "X happens when Y" or just "X happens."
SY-03: Passive Voice Overuse¶
PATTERN: "The report was generated," "The issue was resolved," "The data was processed" FIX: name the actor. "The agent generated the report." Or imperative: "Generate the report."
SY-04: Adverb Stuffing¶
PATTERN: "significantly improves," "dramatically reduces," "fundamentally changes," "effectively manages" FIX: cut the adverb. If the verb needs an adverb to have impact, pick a stronger verb. BEFORE: "This significantly reduces deployment time." AFTER: "This cuts deployment time by 60%."
SY-05: Em-Dash Overuse¶
PATTERN: 3+ em-dashes per paragraph — used for every aside — creating a choppy reading experience FIX: max 1 em-dash per paragraph. Use parentheses, commas, or separate sentences for other asides.
SY-06: Colon-List Pattern¶
PATTERN: "There are several key aspects to consider:" followed by a list. Repeats every section. FIX: vary how you introduce information. Don't announce lists — just start them.
SY-07: Sycophantic Adjectives¶
PATTERN: "excellent," "remarkable," "outstanding," "impressive" describing the subject's own work FIX: let the facts speak. Numbers beat adjectives. "97% uptime" > "remarkable reliability"
SY-08: Equal-Weight Enumeration¶
PATTERN: every item in a list gets the same amount of detail, even when importance varies wildly FIX: spend more words on what matters most. One-line some items. Deep-dive others.
CATEGORY_4: FILLER_PATTERNS¶
Deduction: -1 point each occurrence
FL-01: "In today's [adjective] [noun]" Openings¶
PATTERN: "In today's fast-paced digital landscape," "In today's competitive market," FIX: delete the entire opening. Start with your actual point.
FL-02: "This comprehensive guide"¶
PATTERN: self-referential meta-description ("This article explores," "In this guide, we'll cover") FIX: just start writing the content. The reader knows what they're reading.
FL-03: "Let's explore" / "Let's dive in"¶
PATTERN: fake-collaborative invitation to read the next section FIX: delete. The reader is already reading.
FL-04: "At the end of the day"¶
PATTERN: AI crutch phrase for "ultimately" or "the bottom line" FIX: say what you mean directly, or cut the sentence entirely.
FL-05: Restating the Obvious¶
PATTERN: "As we all know, security is important." "It goes without saying that quality matters." FIX: if it goes without saying, don't say it.
FL-06: "Whether you're a [X] or a [Y]"¶
PATTERN: false-inclusive audience addressing. "Whether you're a startup founder or an enterprise CTO..." FIX: know your audience. Write for them specifically.
CATEGORY_5: COMMUNICATION_TELLS¶
Deduction: -2 points each occurrence
CT-01: Apologetic Deflection¶
PATTERN: "I can't do X, but here's what I can do" structure in non-interactive copy FIX: this is conversational AI leakage into written copy. Written copy doesn't apologize — it states.
CT-02: Bullet-Point Walls¶
PATTERN: 10+ consecutive bullet points where a paragraph would be clearer FIX: group into 3-4 bullets max, then prose. Or use a table. Long bullet lists are AI's comfort zone.
CT-03: "Key Takeaways" Boxes¶
PATTERN: every article ends with "Key Takeaways:" followed by a bulleted summary FIX: the article IS the takeaway. If readers need a summary, the article was too long.
CT-04: Rhetorical Question + Immediate Answer¶
PATTERN: "But what makes X different? The answer lies in Y." FIX: just state Y. The rhetorical question adds nothing.
CT-05: Grandiose Claims Without Evidence¶
PATTERN: "This changes everything." "The future of X is here." FIX: show, don't tell. Specific numbers, specific examples, specific results.
CT-06: Corporate "We" Without Personality¶
PATTERN: "We believe in delivering excellence." "We're committed to innovation." FIX: say something specific. "We run 59 agents that ship production code." Facts, not mission statements.
REMEDIATION_PROTOCOL¶
STEP 1: complete draft — finish the full piece before scanning STEP 2: run self-check — scan against all 5 categories above STEP 3: count deductions — track by category STEP 4: IF score < 85 THEN rewrite flagged sections STEP 5: IF score < 70 THEN fresh redraft (keep the brief, rewrite from scratch) STEP 6: re-scan after fixes — common mistake: fixing one pattern while introducing another STEP 7: submit to Leah for expert panel scoring (AI Humanizer dimension at 1.5x weight)
REWRITE_TECHNIQUE¶
RULE: read the flagged sentence aloud — if it sounds like a press release, rewrite RULE: shorter is almost always better — cut 20% after every pass RULE: add specifics — replace every adjective with a number or example RULE: vary your rhythm — alternate short and long sentences RULE: one idea per sentence — if you need a semicolon, make it two sentences
BEFORE_AFTER_EXAMPLES¶
Vocabulary + Filler Fix¶
BEFORE: "In today's rapidly evolving digital landscape, businesses need to leverage cutting-edge AI solutions to streamline their operations and drive transformative results." AFTER: "Most businesses waste 40% of developer time on repetitive tasks. AI agents handle those tasks at 10% of the cost." SCORE_CHANGE: -2 (landscape) -2 (leverage) -2 (cutting-edge) -2 (streamline) -2 (transformative) -1 (In today's) = -11 → rewritten: 0 deductions
Structural Fix¶
BEFORE:
AI agents offer three key advantages:
- They automate repetitive workflows
- They improve code quality
- They reduce development costs
Furthermore, these agents provide three additional benefits:
- They learn from every project
- They maintain consistent standards
- They scale without hiring
The automation is the obvious win — our agents handle deployment pipelines,
test generation, and status updates without anyone asking.
What surprised us was the consistency. Agent #47 (Julian, our compliance officer)
catches the same edge cases at 2 AM that he catches at 2 PM. Human reviewers
don't do that.
Scaling is just math after that. One agent costs €0.03/task. One developer
costs €45/hour.
Style Fix¶
BEFORE: "It's worth noting that our innovative approach significantly reduces the time needed to effectively manage complex software deployments." AFTER: "Deployments that took 4 hours now take 12 minutes." SCORE_CHANGE: -2 (worth noting) -2 (innovative) -2 (significantly) -2 (effectively) = -8 → rewritten: 0 deductions
READ_ALSO: domains/creative/copywriting-craft.md READ_ALSO: domains/creative/brand-voice.md READ_ALSO: domains/creative/expert-panel-scoring.md READ_ALSO: domains/marketing/content-marketing.md