Writing Standards Structured Copy-Editing — Knowledge Module Reference

Copy-Editing

A structured editing methodology for reviewing and polishing existing marketing copy. Uses the Nine Sweeps framework to systematically improve clarity, persuasion, credibility, and readability in separate focused passes. Each sweep targets one dimension so nothing gets missed and nothing gets over-edited.

This module is brand-agnostic. It works on any marketing content: landing pages, emails, blog posts, ad copy, product descriptions, case studies. Brand-specific voice and constraints come from the consumer's brand-context.md file.

Brand Context

Read brand-context.md before editing any content. You need to understand the intended voice so you preserve it. Editing that strips brand personality is worse than leaving rough copy alone.

Key things to absorb from brand context before you start:

• Tone and formality — is this brand casual, professional, technical, playful?
• Approved language — terms the brand uses deliberately, even if they seem unusual
• Banned language — words or phrases the brand avoids for strategic reasons
• Audience — who reads this? What do they already know? What do they care about?

If no brand context is available, ask for one. Without it, you risk editing toward generic corporate voice.

The Nine Sweeps

The core methodology. Each sweep is a single focused pass through the content targeting one specific dimension. Do not combine sweeps. Do not skip sweeps. The order matters: structural issues first, polish last.

Sweep 1: Clarity

Focus: Can the reader understand this on first read?

What to check:

• Sentences that require re-reading to parse
• Unclear pronoun references ("it", "this", "they" without obvious antecedent)
• Jargon or acronyms used without explanation
• Ambiguous modifiers (does the adjective apply to the first noun or the second?)
• Sentences trying to say two things at once

Common issues:

• Stacking multiple clauses in one sentence
• Using "this" to refer to an entire preceding paragraph
• Assuming the reader shares context they don't have
• Leading with qualifications before the main point

Process:

1. Read each sentence in isolation. Does it make sense without the surrounding context?
2. Identify every pronoun. Can you point to exactly what it refers to?
3. Flag any sentence longer than 30 words. Can it be split?
4. Mark any term that would confuse a new reader of this content.
5. Check opening sentences of each section. Do they orient the reader?

After this sweep: every sentence should be parseable on first read by someone in the target audience.

---

Sweep 2: Voice & Tone

Focus: Does every sentence sound like the same author?

What to check:

• Formality shifts within the same piece (casual in one paragraph, corporate in the next)
• Personality breaks (warm and direct, then suddenly stiff and distant)
• Register inconsistency (contractions in one spot, formal constructions in another)
• Borrowed voice patterns (sounds like a different brand or writer mid-piece)

Common issues:

• Opening paragraphs written in one voice, body copy in another (common when multiple people contributed)
• CTAs that suddenly shift to aggressive sales tone when the rest is conversational
• Technical sections that go cold and impersonal when the rest is warm
• Hedging language that contradicts a confident brand voice

Process:

1. Read the first paragraph aloud (mentally). Fix the voice in your head.
2. Read each subsequent paragraph. Does it sound like the same person?
3. Check all CTAs. Do they match the tone of the surrounding content?
4. Look for formality mismatches: "don't" next to "do not", "get" next to "acquire."
5. Verify against brand context: does this match the documented voice?

After this sweep: the piece should read as though one consistent person wrote every word.

---

Sweep 3: So What

Focus: Does every paragraph answer "so what?" for the reader?

What to check:

• Features described without connecting to reader benefit
• Claims made without explaining why the reader should care
• Paragraphs that inform but don't motivate
• Missing "which means..." bridges between fact and value
• Sections that describe what something is but not what it does for the reader

Common issues:

• Leading with the technology or process instead of the outcome
• Listing capabilities without tying them to pain points
• Telling the brand's story instead of the reader's story
• Assuming the reader will connect the dots on their own

Process:

1. Read each paragraph and ask: "If I'm the reader, why do I care about this?"
2. For every feature mentioned, check: is the benefit stated or implied?
3. For every claim, check: is there a "which means..." connection to the reader's world?
4. Highlight any paragraph that only describes. Add the "so what" bridge.
5. Check headings: do they promise something the reader wants, or just label content?

After this sweep: every paragraph should give the reader a reason to keep reading.

---

Sweep 4: Prove It

Focus: Would a skeptic accept this?

What to check:

• Unsubstantiated claims ("industry-leading", "best-in-class", "revolutionary")
• Missing social proof where it would strengthen the argument
• Statistics without sources or context
• Promises without evidence or mechanism
• Superlatives that invite doubt

Common issues:

• Claiming results without citing a customer, study, or data point
• Using social proof that's vague ("trusted by thousands" instead of "used by 2,400 teams")
• Making comparison claims without naming what you're comparing to
• Stacking adjectives instead of providing evidence

Process:

1. Underline every claim that a reader could challenge with "prove it."
2. For each one: is there evidence in the copy? A number, a quote, a case study reference?
3. If no evidence exists, either add it, soften the claim, or remove it.
4. Check all numbers: are they specific enough? "Over 100" is weaker than "140+."
5. Verify that social proof is concrete: named companies, real quotes, specific outcomes.

After this sweep: every claim should be backed by evidence, or stated as an opinion with appropriate hedging.

---

Sweep 5: Specificity

Focus: Can I replace any vague word with a specific one?

What to check:

• Vague quantifiers ("many", "several", "various", "numerous")
• Round numbers that feel estimated ("about 100" vs. "97")
• Abstract nouns that could be concrete ("solution", "platform", "tool")
• Generic descriptions ("high-quality", "user-friendly", "powerful")
• Missing details that would make the content more credible

Common issues:

• Using "fast" instead of "under 2 seconds"
• Saying "easy to use" instead of describing what makes it easy
• Writing "saves time" instead of "cuts onboarding from 3 weeks to 4 days"
• Defaulting to category nouns ("our platform") instead of naming specific capabilities

Process:

1. Highlight every adjective and adverb. Can it be replaced with a concrete detail?
2. Find every vague quantifier. Can you insert a real number?
3. Check every "it" and "this" — can you name the specific thing?
4. Look for abstract benefit claims. Can you describe the mechanism?
5. Review metaphors and analogies. Are they concrete enough to visualize?

After this sweep: the copy should feel grounded and precise, not floaty and generic.

---

Sweep 6: Heightened Emotion

Focus: Is there at least one moment that creates tension, relief, or curiosity?

What to check:

• Flat, information-only paragraphs that never engage the reader emotionally
• Missing contrast between the problem and the solution
• No tension-and-release pattern (problem agitation followed by resolution)
• Absence of curiosity gaps (nothing that makes the reader want to know more)
• CTAs that are purely logical without any emotional pull

Common issues:

• Describing the problem without making the reader feel it
• Jumping to the solution before adequately agitating the pain
• Treating every paragraph with the same emotional weight
• Using rational arguments exclusively when the audience decides emotionally

Process:

1. Map the emotional arc of the piece. Does it go anywhere, or is it flat?
2. Find the problem statement. Does it make the reader feel the pain, or just understand it?
3. Check for at least one contrast: before/after, with/without, old way/new way.
4. Look for curiosity gaps: does the copy ever make the reader lean in?
5. Check the CTA: does it connect to a desire, not just a feature?

After this sweep: the copy should have emotional peaks and valleys, not monotone information delivery.

---

Sweep 7: Zero Risk

Focus: Is there anything that could make someone hesitate?

What to check:

• Friction near CTAs (confusing next steps, unclear commitment level)
• Unanswered objections that the reader is likely thinking
• Missing trust signals (guarantees, security badges, privacy assurances)
• Ambiguous pricing or commitment language
• Any sentence that might trigger buyer anxiety

Common issues:

• CTA buttons that don't clarify what happens next ("Submit" vs. "Start your free trial")
• No mention of cancellation policy, refund terms, or data handling
• Missing answers to "What if it doesn't work for me?"
• Free trials that don't emphasize "no credit card required" (if applicable)
• Jargon near conversion points that makes readers feel unqualified

Process:

1. Read from the perspective of someone looking for reasons NOT to act.
2. List every objection a reasonable person might have at each stage.
3. Check: does the copy address these objections near where they arise?
4. Review every CTA: is the next step crystal clear? Is commitment level stated?
5. Look for risk reversals: guarantees, trials, easy exit. Are they visible enough?

After this sweep: a reader should feel that acting is lower risk than not acting.

---

Sweep 8: Data Verification

Focus: Can every claim be backed up?

What to check:

• Every metric is traceable to a source document, study, or internal data
• No numbers that can't be verified if challenged
• Relative metrics ("50% faster") state the baseline they're compared against
• Date-sensitive data is current, not stale
• Percentages and multiples are calculated correctly

Common issues:

• Metrics pulled from outdated reports
• Percentages without base numbers ("90% of users" — 90% of how many?)
• Competitor comparisons using different methodologies
• Rounding that makes numbers misleading
• Growth metrics without timeframes ("grew 300%" — over what period?)

Process:

1. List every number, statistic, and quantitative claim in the copy.
2. For each one, identify the source. If you can't find one, flag it.
3. Check that relative claims state their baseline ("faster than what?").
4. Verify date-sensitivity: is this data still current?
5. Confirm that any "up to" or "as much as" claims have real-world backing.

After this sweep: every number in the copy should have a traceable source, and every comparison should be fair.

---

Sweep 9: Anti-AI-Slop

Focus: Would this pass an AI detection check?

If the check_writing_quality tool is available (writing-quality plugin), run it. Otherwise, manually check the patterns below.

The plugin provides:

• Deterministic pattern matching for banned phrases, structures, and AI vocabulary (instant, no LLM)
• LLM-based 5-dimension scoring: directness, rhythm, trust, authenticity, density
• Threshold: 35/50 to pass. Any single dimension below 5 triggers revision.

What to check:

• Rule-of-three lists (two items beat three — three is one of the strongest AI tells)
• Em dash overuse (replace with commas, periods, or parentheses)
• Significance inflation ("revolutionary", "game-changing", "transformative")
• Sycophantic or chatbot artifacts ("Great question!", "I hope this helps!")
• Formulaic structures: binary contrasts ("Not X. But Y."), dramatic fragmentation, false agency

Common issues:

• Copy that passed all previous sweeps but still reads as machine-generated
• Overly smooth transitions (real writing has some roughness)
• Every paragraph ending with a punchy one-liner (that's a pattern, not style)
• Synonym cycling to avoid repetition (using four different words for the same thing)

Process:

1. Run check_writing_quality tool in full mode (or patterns-only for fast feedback).
2. If plugin unavailable, manually check for: filler phrases, dramatic fragmentation, rule of three, passive voice, em dashes, binary contrasts, AI vocabulary, negative listing.
3. Flag all detected patterns with category and severity.
4. Rewrite flagged sections: cut filler, use active voice, name actors, vary rhythm, be specific.
5. Re-run quality check after revision. Confirm 35/50 threshold met.

After this sweep: the copy should sound like a specific human wrote it, not like AI generated it.

---

Quick-Pass Editing Checks

When you don't have time for the full Nine Sweeps, run these rapid checks at three levels:

Word Level

• Weak verbs — replace "is", "has", "makes" with verbs that carry meaning ("drives", "eliminates", "unlocks")
• Adverbs — cut most -ly words. "Runs quickly" becomes "runs fast" or just pick a stronger verb
• Filler words — cut "very", "really", "actually", "basically", "just", "quite", "simply"
• Jargon — replace complex words with plain alternatives (see references/plain-english-alternatives.md)
• Redundancies — "free gift", "advance planning", "end result", "past history"

Sentence Level

• Passive voice — name the actor. "Errors are reduced" becomes "The system catches errors."
• Sentence length — flag anything over 30 words. Split or tighten.
• Repeated starts — three sentences in a row starting the same way (especially "We", "Our", "The", "It")
• Throat-clearing — delete opening clauses that delay the point ("It's important to note that", "As you may know")
• Buried leads — the most important word in the sentence should be near the beginning or end, not in the middle

Paragraph Level

• Missing transitions — does each paragraph connect logically to the next?
• Length variation — mix short paragraphs (1-2 sentences) with longer ones (3-4). Never exceed 5 sentences.
• Purpose — can you state what each paragraph does in one word? (hooks, explains, proves, converts) If you can't, the paragraph is unfocused.
• One idea — each paragraph should contain one idea. If it contains two, split it.

Common Copy Problems & Fixes

Problem	Example	Fix
Feature-first writing	"Our platform uses ML-powered algorithms"	"Cut analysis time from 3 hours to 10 minutes" (lead with outcome)
Vague benefits	"Save time and money"	"Teams save an average of 12 hours per week on manual reporting"
Stacked adjectives	"Our powerful, innovative, industry-leading solution"	Pick one. Back it with evidence.
Weak CTAs	"Learn More"	"Start your free trial" or "See it in action" (specific next step)
Passive credibility	"Trusted by leading companies"	"Used by 340 teams including Stripe, Notion, and Linear"
Buried value prop	"Founded in 2019, our team of 50 engineers built a platform that..."	Lead with what the reader gets. Company backstory goes below the fold.
Hedging language	"We believe our product can potentially help..."	"Our product reduces churn by 23%." State it or don't.
Walls of text	Five consecutive 6-sentence paragraphs	Break into scannable chunks: subheadings, short paragraphs, bullet points
Inconsistent capitalization	"Machine Learning" in one place, "machine learning" in another	Pick a style, document it, apply it everywhere
Missing proof	"The fastest solution on the market"	"Processes 10K records in 1.2 seconds (3x faster than the next alternative)"

Common Tasks

1. Run full Nine Sweeps — execute all nine sweeps in order on the provided content. Return marked-up copy with changes annotated by sweep number.
2. Run a targeted single sweep — run only one specific sweep (e.g., "run Sweep 4: Prove It on this landing page"). Useful when you know the weak area.
3. Quick-pass check — run word/sentence/paragraph-level checks without the full sweep methodology. For fast feedback during drafting.
4. Data verification only — Sweep 8 in isolation. Audit every number, metric, and quantitative claim. Return a table of claims with source status.
5. Anti-AI-slop only — Sweep 9 in isolation. Run check_writing_quality tool (or manual checks) and return pattern detection results and scoring.

Tips

• Edit in passes, not all at once. Trying to fix clarity, voice, and evidence simultaneously means you fix none of them well.
• Read the copy as a first-time visitor, not as someone who already knows the product. The curse of knowledge is the number one cause of unclear copy.
• Preserve the writer's voice. Your job is to make their writing better, not replace it with yours.
• When in doubt, cut. Shorter copy almost always outperforms longer copy. Every word should earn its place.
• Track the changes you make most often. If you're always fixing the same issue, that's a process problem upstream, not an editing problem.
• After editing, re-read the whole piece start to finish. Individual sentence fixes can break paragraph flow.
• Use references/plain-english-alternatives.md when simplifying language. Plain words beat complex ones.
• Use references/quality-checklist.md as a final sign-off before delivery.

Gotchas

• Over-editing — the biggest risk. Aggressive editing can strip personality, flatten voice, and produce technically correct but lifeless copy. After each sweep, re-read to make sure you haven't homogenized the writing.
• Voice stripping — editing for clarity can accidentally remove deliberate stylistic choices. Check brand context before changing anything that looks like a voice choice (sentence fragments, colloquialisms, unconventional punctuation).
• Sweep creep — don't fix Sweep 5 problems during Sweep 2. Stay in your lane per sweep, or you'll miss things and introduce inconsistency.
• Perfectionism — diminishing returns hit hard after Sweep 7. If the copy is clear, on-voice, persuasive, and credible, minor word-level polishing rarely moves the needle.
• Audience mismatch — editing B2B enterprise copy with the same standards as consumer copy (or vice versa) produces wrong results. Calibrate to the audience.
• Data fabrication — during Sweep 4 (Prove It), never invent statistics or social proof to fill gaps. Flag the gap and ask for real data.

Related Modules

• copywriting (writing/copywriting/_skill.md) — for drafting new content from scratch. Use copywriting to write, copy-editing to review.
• writing-quality plugin (check_writing_quality MCP tool) — AI pattern detection and scoring engine. Install separately from sidebutton/plugin-writing-quality. Called during Sweep 9.
• content-strategy (writing/content-strategy/_skill.md) — for planning what content to create. Ensures edited content aligns with broader content goals.